)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":29577,"name":"Parav Pandit","email":"parav@mellanox.com"},"change_message_id":"daf57be8a59b30e3c2b228712d275f7cb95af7d0","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Add spec for sriov live migration"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"- This change introduces a spec to describe"},{"line_number":10,"context_line":"  extention of the nova libvirt dirver to"},{"line_number":11,"context_line":"  support live migration of guests with"},{"line_number":12,"context_line":"  SR-IOV interfaces."},{"line_number":13,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":6,"id":"3f79a3b5_492f6ea5","line":10,"updated":"2018-11-26 22:18:23.000000000","message":"driver*","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Add spec for sriov live migration"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"- This change introduces a spec to describe"},{"line_number":10,"context_line":"  extention of the nova libvirt driver to"},{"line_number":11,"context_line":"  support live migration of guests with"},{"line_number":12,"context_line":"  SR-IOV interfaces."},{"line_number":13,"context_line":""}],"source_content_type":"text/x-gerrit-commit-message","patch_set":10,"id":"dfd5e7cf_aa4a0852","line":10,"range":{"start_line":10,"start_character":2,"end_line":10,"end_character":11},"updated":"2019-01-10 14:55:52.000000000","message":"extension","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"}],"specs/stein/approved/libvirt-neutron-sriov-livemigration.rst":[{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"0ecd646743aa1dc8b7693895163974a00a461bdf","unresolved":false,"context_lines":[{"line_number":1,"context_line":".."},{"line_number":2,"context_line":""},{"line_number":3,"context_line":" This work is licensed under a Creative Commons Attribution 3.0 Unported"},{"line_number":4,"context_line":""},{"line_number":5,"context_line":" License."},{"line_number":6,"context_line":""},{"line_number":7,"context_line":" http://creativecommons.org/licenses/by/3.0/legalcode"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Neutron SR-IOV Port LiveMigration"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":""},{"line_number":16,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":17,"context_line":""},{"line_number":18,"context_line":""},{"line_number":19,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":20,"context_line":""},{"line_number":21,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":26,"context_line":""},{"line_number":27,"context_line":"seeks to address the gaps in live migration support."},{"line_number":28,"context_line":""},{"line_number":29,"context_line":""},{"line_number":30,"context_line":"Problem description"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":33,"context_line":""},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Live Migration with SR-IOV devices has several complicating factors."},{"line_number":36,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3f79a3b5_5e7a57f0","line":33,"range":{"start_line":1,"start_character":0,"end_line":33,"end_character":0},"updated":"2018-10-10 10:48:52.000000000","message":"Not sure what all the newlines are doing here. Typo?","commit_id":"03f3e07f6de24560630213252b022354fb3e61e8"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":14,"context_line":""},{"line_number":15,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":18,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":19,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":20,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":21,"context_line":"seeks to address the gaps in live migration support."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_020cbd36","line":18,"range":{"start_line":17,"start_character":0,"end_line":18,"end_character":41},"updated":"2018-10-04 14:37:44.000000000","message":"nit: Could we invert the cause/effect order, ie \"Ever since SRIOV was added, ...\"","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":14,"context_line":""},{"line_number":15,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":18,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":19,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":20,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":21,"context_line":"seeks to address the gaps in live migration support."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_298e207a","line":18,"range":{"start_line":17,"start_character":0,"end_line":18,"end_character":41},"in_reply_to":"3f79a3b5_020cbd36","updated":"2018-10-09 17:11:05.000000000","message":"sure ill repharse","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":16,"context_line":""},{"line_number":17,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":18,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":19,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":20,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":21,"context_line":"seeks to address the gaps in live migration support."},{"line_number":22,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_c211e551","line":19,"range":{"start_line":19,"start_character":44,"end_line":19,"end_character":53},"updated":"2018-10-04 14:37:44.000000000","message":"\"migrate\"","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":17,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":18,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":19,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":20,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":21,"context_line":"seeks to address the gaps in live migration support."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_6226d1bc","line":20,"range":{"start_line":20,"start_character":31,"end_line":20,"end_character":42},"updated":"2018-10-04 14:37:44.000000000","message":"maybe a quick elaboration as why it\u0027s not correct?","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":17,"context_line":"SR-IOV live migration has always been an edge case since"},{"line_number":18,"context_line":"support for SR-IOV was added to OpenStack. With vNic type"},{"line_number":19,"context_line":"macvtap it is functionally possible to live migration instances"},{"line_number":20,"context_line":"today with OpenStack but it is not correct to do so. This spec"},{"line_number":21,"context_line":"seeks to address the gaps in live migration support."},{"line_number":22,"context_line":""},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_099324a6","line":20,"range":{"start_line":20,"start_character":31,"end_line":20,"end_character":42},"in_reply_to":"3f79a3b5_6226d1bc","updated":"2018-10-09 17:11:05.000000000","message":"well i didn\u0027t want to go into it in the intro but because we don\u0027t claim the device on the destination host.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":32,"context_line":"*     resource claims"},{"line_number":33,"context_line":""},{"line_number":34,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":35,"context_line":"separately by `[1]`"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":38,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_bf23a2ba","line":35,"range":{"start_line":35,"start_character":14,"end_line":35,"end_character":19},"updated":"2018-10-04 14:37:44.000000000","message":"nit: need a tail (underscore) here for the reference to work.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"0ecd646743aa1dc8b7693895163974a00a461bdf","unresolved":false,"context_lines":[{"line_number":32,"context_line":"*     resource claims"},{"line_number":33,"context_line":""},{"line_number":34,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":35,"context_line":"separately by `[1]`"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":38,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_f20d5ae7","line":35,"range":{"start_line":35,"start_character":14,"end_line":35,"end_character":19},"in_reply_to":"3f79a3b5_29c6206b","updated":"2018-10-10 10:48:52.000000000","message":"You should also drop the backticks","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":32,"context_line":"*     resource claims"},{"line_number":33,"context_line":""},{"line_number":34,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":35,"context_line":"separately by `[1]`"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":38,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_29c6206b","line":35,"range":{"start_line":35,"start_character":14,"end_line":35,"end_character":19},"in_reply_to":"3f79a3b5_bf23a2ba","updated":"2018-10-09 17:11:05.000000000","message":"oh ok. thats a little annoying but if its needed i guess ill add it :)","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":38,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":39,"context_line":"state. NOTE: hardware state in this context refers to any nic state such as"},{"line_number":40,"context_line":"enabled offloads or Tx/Rx queues that are implemented in hardware which is not"},{"line_number":41,"context_line":"software programmable via the hypervisor e.g. mac address is not considered"},{"line_number":42,"context_line":"hardware state in this context as libvirt/qemu can set the mac of an SR-IOV"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_828bed1b","line":39,"range":{"start_line":39,"start_character":7,"end_line":39,"end_character":11},"updated":"2018-10-04 14:37:44.000000000","message":"If we\u0027re going to do a NOTE might as well do it \"properly\" with rst.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":38,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":39,"context_line":"state. NOTE: hardware state in this context refers to any nic state such as"},{"line_number":40,"context_line":"enabled offloads or Tx/Rx queues that are implemented in hardware which is not"},{"line_number":41,"context_line":"software programmable via the hypervisor e.g. mac address is not considered"},{"line_number":42,"context_line":"hardware state in this context as libvirt/qemu can set the mac of an SR-IOV"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_49eefc21","line":39,"range":{"start_line":39,"start_character":7,"end_line":39,"end_character":11},"in_reply_to":"3f79a3b5_828bed1b","updated":"2018-10-09 17:11:05.000000000","message":"sure i have seen that done before i just could not remember the syntax ill look it up","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":43,"context_line":"device via a standard host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. this spec will focus on 2 categories primarily on direct"},{"line_number":47,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":48,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":49,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_428155f9","line":46,"range":{"start_line":46,"start_character":13,"end_line":46,"end_character":17},"updated":"2018-10-04 14:37:44.000000000","message":"This","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":43,"context_line":"device via a standard host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. this spec will focus on 2 categories primarily on direct"},{"line_number":47,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":48,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":49,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_a2714947","line":46,"range":{"start_line":46,"start_character":50,"end_line":46,"end_character":59},"updated":"2018-10-04 14:37:44.000000000","message":"nit: don\u0027t think we need this word, ie \"categories: direct passthrough (...) and indirect passthrough.\"","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":43,"context_line":"device via a standard host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. this spec will focus on 2 categories primarily on direct"},{"line_number":47,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":48,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":49,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_29414009","line":46,"range":{"start_line":46,"start_character":50,"end_line":46,"end_character":59},"in_reply_to":"3f79a3b5_a2714947","updated":"2018-10-09 17:11:05.000000000","message":"well i said primarialy as i do refer to virtio-forwarder which will be enabled as a secondary effect of enableing indirect passthrough. similarly hardware offloaded ovs will be enabled as a secondary effect of direct passthough.\n\nin both cases addressing them directly in the spec add complexity to the discussion that is not required to discus the change. they will be fixed for free by fixing generic sriov.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":48,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":49,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"when a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":52,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":53,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":54,"context_line":"state direct mode SR-IOV cannot be conventionally live migrated."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_628d7128","line":51,"range":{"start_line":51,"start_character":0,"end_line":51,"end_character":4},"updated":"2018-10-04 14:37:44.000000000","message":"When","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":58,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":59,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed."},{"line_number":61,"context_line":"Indirect mode SR-IOV therefor allows migration of guests."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_8290cd43","line":61,"range":{"start_line":61,"start_character":21,"end_line":61,"end_character":29},"updated":"2018-10-04 14:37:44.000000000","message":"therefore","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do so resources are not correctly claimed and migration will fail"},{"line_number":67,"context_line":"if the host PCI address of the SR-IOV device is not free on the destination"},{"line_number":68,"context_line":"node. This is because of 2 issues. First we do not correctly claim the"},{"line_number":69,"context_line":"SR-IOV device on the destination node and Second, we do not modify the guest"},{"line_number":70,"context_line":"XML to reflect the host PCI address on the destination."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_2256995f","line":67,"range":{"start_line":67,"start_character":7,"end_line":67,"end_character":56},"updated":"2018-10-04 14:37:44.000000000","message":"I get what you\u0027re saying, but it could probably be made more explicit - either there\u0027s a PCI device on the dest with the same address as the host, *and* it\u0027s an SRIOV card *and* it\u0027s free (only successful case), or it\u0027s not an SRIOV card (fail), or it\u0027s an SRIOV card and it\u0027s not free (fail), or a device with that address doesn\u0027t even exist on the dest (fail). Did I get that right?","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do so resources are not correctly claimed and migration will fail"},{"line_number":67,"context_line":"if the host PCI address of the SR-IOV device is not free on the destination"},{"line_number":68,"context_line":"node. This is because of 2 issues. First we do not correctly claim the"},{"line_number":69,"context_line":"SR-IOV device on the destination node and Second, we do not modify the guest"},{"line_number":70,"context_line":"XML to reflect the host PCI address on the destination."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_690df8e5","line":67,"range":{"start_line":67,"start_character":7,"end_line":67,"end_character":56},"in_reply_to":"3f79a3b5_2256995f","updated":"2018-10-09 17:11:05.000000000","message":"yes exactly that just felt a little long to directly call out but i can add it.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":71,"context_line":""},{"line_number":72,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":73,"context_line":"currently unreliable and incorrect when it succeeds."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"Use Cases"},{"line_number":77,"context_line":"---------"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_a2de29d8","line":74,"updated":"2018-10-04 14:37:44.000000000","message":"Super good problem description, by the way.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":79,"context_line":"As a telecom operator with stateful VNF such as a vPE Router"},{"line_number":80,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":81,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":82,"context_line":"flexibility for migration for maintenance. To that end, as an operator,"},{"line_number":83,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":84,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":85,"context_line":"understanding performance SLAs will not be met during the migration."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_82968dee","line":82,"range":{"start_line":82,"start_character":16,"end_line":82,"end_character":25},"updated":"2018-10-04 14:37:44.000000000","message":"Should we replace \"migration\" with \"live-migration\" everywhere in this spec? Cold migration exists as well, but that\u0027s not what the telcos want, although from context it\u0027s probably obvious that we\u0027re always talking about live migration.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":79,"context_line":"As a telecom operator with stateful VNF such as a vPE Router"},{"line_number":80,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":81,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":82,"context_line":"flexibility for migration for maintenance. To that end, as an operator,"},{"line_number":83,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":84,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":85,"context_line":"understanding performance SLAs will not be met during the migration."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_62c57109","line":82,"range":{"start_line":82,"start_character":12,"end_line":82,"end_character":15},"updated":"2018-10-04 14:37:44.000000000","message":"nit: \"of\"?","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":79,"context_line":"As a telecom operator with stateful VNF such as a vPE Router"},{"line_number":80,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":81,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":82,"context_line":"flexibility for migration for maintenance. To that end, as an operator,"},{"line_number":83,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":84,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":85,"context_line":"understanding performance SLAs will not be met during the migration."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_a99f9062","line":82,"range":{"start_line":82,"start_character":16,"end_line":82,"end_character":25},"in_reply_to":"3f79a3b5_82968dee","updated":"2018-10-09 17:11:05.000000000","message":"i can be explcit and say live migration that said i was making that substituion in my head when reading/writing based on context.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":98,"context_line":"Resource claims"},{"line_number":99,"context_line":"---------------"},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"building on top of the recently added multiple port binding feature this"},{"line_number":102,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":103,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":104,"context_line":"tracker. if claiming fails then the partially claimed resources will be"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_c287a5b2","line":101,"range":{"start_line":101,"start_character":0,"end_line":101,"end_character":8},"updated":"2018-10-04 14:37:44.000000000","message":"Building","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":101,"context_line":"building on top of the recently added multiple port binding feature this"},{"line_number":102,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":103,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":104,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":105,"context_line":"released and check_can_live_migrate_destination will fail. if the claiming"},{"line_number":106,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":107,"context_line":"SR-IOV devices will be updated with the dest host PCI address."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_42b5d57b","line":104,"range":{"start_line":104,"start_character":9,"end_line":104,"end_character":11},"updated":"2018-10-04 14:37:44.000000000","message":"If","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":102,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":103,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":104,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":105,"context_line":"released and check_can_live_migrate_destination will fail. if the claiming"},{"line_number":106,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":107,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":108,"context_line":"if the migration should fail after the destination resources"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_a2a5c949","line":105,"range":{"start_line":105,"start_character":59,"end_line":105,"end_character":61},"updated":"2018-10-04 14:37:44.000000000","message":"If - OK, I think you know first words of sentences should be capitalized, I\u0027ll stop now ;)","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":102,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":103,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":104,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":105,"context_line":"released and check_can_live_migrate_destination will fail. if the claiming"},{"line_number":106,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":107,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":108,"context_line":"if the migration should fail after the destination resources"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_c999ac7a","line":105,"range":{"start_line":105,"start_character":59,"end_line":105,"end_character":61},"in_reply_to":"3f79a3b5_a2a5c949","updated":"2018-10-09 17:11:05.000000000","message":"yes but my text editor does not and does not auto correect this :)","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":108,"context_line":"if the migration should fail after the destination resources"},{"line_number":109,"context_line":"have been claimed they must be released in the"},{"line_number":110,"context_line":"rollback_live_migration_at_destination call."},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"XML Generation"},{"line_number":113,"context_line":"--------------"},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_9301fd09","line":111,"updated":"2018-10-04 14:37:44.000000000","message":"For NUMA live migration we don\u0027t really have a choice in using a MigrationContext because that\u0027s the only way the resource tracker can handle MoveClaims. Would it make sense to re-use the same mechanism here?\n\nPreliminary, I think the context/claim would have to happen at the conductor, because they need DB access, and computes don\u0027t have it. There\u0027s no indirection like with objects.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":108,"context_line":"if the migration should fail after the destination resources"},{"line_number":109,"context_line":"have been claimed they must be released in the"},{"line_number":110,"context_line":"rollback_live_migration_at_destination call."},{"line_number":111,"context_line":""},{"line_number":112,"context_line":"XML Generation"},{"line_number":113,"context_line":"--------------"},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_c9720c4b","line":111,"in_reply_to":"3f79a3b5_9301fd09","updated":"2018-10-09 17:11:05.000000000","message":"it might althouhg i had planned on not using MoveClaims.\ninstead i was going to create a second set of claims for the destination in check_can_live_migrate_destination and freeing the source claims in post_live_migration (clean up source) on success and rollback_live_migration_at_destination on failure.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":123,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"Direct Mode"},{"line_number":127,"context_line":"-------------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"For direct mode  SR-IOV to enable live migration the SR-IOV devices must"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_0ed70624","line":126,"updated":"2018-10-04 14:37:44.000000000","message":"Is this... useful? I don\u0027t disagree that it could be done, but if we\u0027re killing the workload\u0027s network connection, is there really any added value over regular cold migration?","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"6e921b97dbca2abce80e3cc2652031da225563c4","unresolved":false,"context_lines":[{"line_number":123,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"Direct Mode"},{"line_number":127,"context_line":"-------------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"For direct mode  SR-IOV to enable live migration the SR-IOV devices must"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_6d4c4f9d","line":126,"in_reply_to":"3f79a3b5_0ed70624","updated":"2018-10-08 04:21:49.000000000","message":"Maybe Sean is talking about a bond at here, I saw that in line 83","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":123,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"Direct Mode"},{"line_number":127,"context_line":"-------------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"For direct mode  SR-IOV to enable live migration the SR-IOV devices must"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_e93ee824","line":126,"in_reply_to":"3f79a3b5_262c1f92","updated":"2018-10-09 17:11:05.000000000","message":"A bond is not strictly required.\nit is need not to have a network interuption during live migration.\n\nEven with a bond live migration implies packet lose but it will hide the removal of the device for application using kernel networking.\n\nthe add value vs cold migrate is with a bound that stateful workloads like PE routers that can take 15+ mins to repeer after cold migrate do not need to rebuild there reouting peers. Without a bound the network interuption can also be short enough to ensure that the peering sessions do not require a rebuild. this will obviously depend on your workload so if you really do not want to loses netwrok state use a bond.\n\ni can state this more explcitly in the spec.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":7543,"name":"Yongli He","email":"yongli.he@intel.com","username":"yongli.he"},"change_message_id":"ed25a13b39272e73c9b265b73cf048c7967bdc36","unresolved":false,"context_lines":[{"line_number":123,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"Direct Mode"},{"line_number":127,"context_line":"-------------"},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"For direct mode  SR-IOV to enable live migration the SR-IOV devices must"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_262c1f92","line":126,"in_reply_to":"3f79a3b5_6d4c4f9d","updated":"2018-10-09 01:07:12.000000000","message":"I guess we need a bond interface here, if so , we may need put here a specific steps/description to avoid misunderstanding.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":198,"context_line":"their config."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"},{"line_number":201,"context_line":"defaults for indirect mode SR-IOV, this feature will be enabled automatically"},{"line_number":202,"context_line":"if both the source and dest node support it."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_2e81a2fa","line":201,"range":{"start_line":201,"start_character":13,"end_line":201,"end_character":21},"updated":"2018-10-04 14:37:44.000000000","message":"You mean direct?","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":198,"context_line":"their config."},{"line_number":199,"context_line":""},{"line_number":200,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"},{"line_number":201,"context_line":"defaults for indirect mode SR-IOV, this feature will be enabled automatically"},{"line_number":202,"context_line":"if both the source and dest node support it."},{"line_number":203,"context_line":""},{"line_number":204,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_e917c89c","line":201,"range":{"start_line":201,"start_character":13,"end_line":201,"end_character":21},"in_reply_to":"3f79a3b5_2e81a2fa","updated":"2018-10-09 17:11:05.000000000","message":"yes","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":213,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":214,"context_line":"however, as that will require both the source and dest node to"},{"line_number":215,"context_line":"this feature, it will have no impact for this release."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Implementation"},{"line_number":218,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":219,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_ce926e58","line":216,"updated":"2018-10-04 14:37:44.000000000","message":"Is there going to be an impact when live migrating between n and n-1 (where n is the version that introduces sriov live migration) compute hosts? Even if the data model doesn\u0027t change, the n-1 host won\u0027t have the necessary code to update the XML.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":213,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":214,"context_line":"however, as that will require both the source and dest node to"},{"line_number":215,"context_line":"this feature, it will have no impact for this release."},{"line_number":216,"context_line":""},{"line_number":217,"context_line":"Implementation"},{"line_number":218,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":219,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_09d6c4c5","line":216,"in_reply_to":"3f79a3b5_ce926e58","updated":"2018-10-09 17:11:05.000000000","message":"just like with the multiple port binding spec if both the source and dest node do not have support we will fall back to old behavior. \n\nSo in this case where n intoduces this behavior in if migrating to n-1 form n we will use the n-1 behavior and when migrating form n-1 to n we will also use the n-1 behavior.\n\nthis significantly reduces the complexity of the code but also means you will only get this functionality if both nodes are running stein or newer.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"5a1f933d7b82a0d9ad44c1f601138da169b2f63b","unresolved":false,"context_lines":[{"line_number":235,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":236,"context_line":""},{"line_number":237,"context_line":"This spec has no dependencies but intends to collaborate with the"},{"line_number":238,"context_line":"implementation of NUMA aware live migration `[1]`"},{"line_number":239,"context_line":""},{"line_number":240,"context_line":"Testing"},{"line_number":241,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_4e8dfe27","line":238,"updated":"2018-10-04 14:37:44.000000000","message":"Still need a tail here :)","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"8558267a1155330143603cd47b018b8235e47eaa","unresolved":false,"context_lines":[{"line_number":235,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":236,"context_line":""},{"line_number":237,"context_line":"This spec has no dependencies but intends to collaborate with the"},{"line_number":238,"context_line":"implementation of NUMA aware live migration `[1]`"},{"line_number":239,"context_line":""},{"line_number":240,"context_line":"Testing"},{"line_number":241,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_e9da68e4","line":238,"in_reply_to":"3f79a3b5_4e8dfe27","updated":"2018-10-09 17:11:05.000000000","message":"yep i think i spotted 2 other cases too.\nill fix them all in the new version.\n\ni proably should have checked the rendered from but we do not warn for this in the docs or pep8 tests which is likely a bug/missing test.","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"0ecd646743aa1dc8b7693895163974a00a461bdf","unresolved":false,"context_lines":[{"line_number":255,"context_line":"References"},{"line_number":256,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":257,"context_line":""},{"line_number":258,"context_line":".. _[1]: https://review.openstack.org/#/c/599587/2/specs/stein/approved/numa-aware-live-migration.rst"},{"line_number":259,"context_line":".. _[2]: https://github.com/openstack/nova/blob/2f635fa914884c91f3b8cc78cda5154dd3b43305/nova/virt/libvirt/driver.py#L2912-L2920"},{"line_number":260,"context_line":".. _[3]: https://github.com/openstack/nova/blob/2f635fa914884c91f3b8cc78cda5154dd3b43305/nova/virt/libvirt/driver.py#L2929-L2932"},{"line_number":261,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3f79a3b5_92cbc671","line":258,"range":{"start_line":258,"start_character":0,"end_line":258,"end_character":8},"updated":"2018-10-10 10:48:52.000000000","message":"Drop the underscore and colon, e.g.\n\n  .. [1] https://...\n\nhttp://rst.ninjs.org/#U2VlIFsxXV8gZm9yIG1vcmUgaW5mb3JtYXRpb24uCgouLiBbMV0gVGVzdA\u003d\u003d","commit_id":"dace1869521d8f90b4048bb5dbc3a938db2a7f66"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":73,"context_line":"pci device on the destination node, with the same pci address as the source"},{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of 2 issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and Second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":81,"context_line":"currently unreliable and incorrect when it succeeds."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_42257c5a","line":78,"range":{"start_line":76,"start_character":0,"end_line":78,"end_character":65},"updated":"2018-10-21 16:30:37.000000000","message":"I believe the latter is only relevant for Direct mode.\n\nIn SR-IOV macvtap, the xml change required is updating the source net device and not the host PCI address.\n(this is already done today)","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":73,"context_line":"pci device on the destination node, with the same pci address as the source"},{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of 2 issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and Second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":81,"context_line":"currently unreliable and incorrect when it succeeds."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_ada53cb6","line":78,"range":{"start_line":76,"start_character":0,"end_line":78,"end_character":65},"in_reply_to":"3f79a3b5_42257c5a","updated":"2018-10-23 13:35:26.000000000","message":"we need to update the pci address in the port binding such that the sriov nic agent can identify the vf correctly but perhaps we do not embed the pci adress of the port used for the macvtap interface in the xml.\n\nfor simplictly i dont want to go into that here.\n\nbut you are correct we may only need to do this for direct mode.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":106,"context_line":"---------------"},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"Building on top of the recently added multiple port binding feature this"},{"line_number":109,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_e2e048fc","line":111,"range":{"start_line":109,"start_character":17,"end_line":111,"end_character":8},"updated":"2018-10-21 16:30:37.000000000","message":"In regards to the PCI resource tracker: will both claim_instance and allocate_instance be done at this point ?","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_c2e50c09","line":113,"range":{"start_line":113,"start_character":32,"end_line":113,"end_character":52},"updated":"2018-10-21 16:30:37.000000000","message":"Did you mean LiveMigrateData object ? if so i believe it\u0027s worth mentioning explicitly.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_42b33c0a","line":113,"range":{"start_line":113,"start_character":13,"end_line":113,"end_character":24},"updated":"2018-10-21 16:30:37.000000000","message":"did you mean VIFMigrateData objects ?","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_adac5c88","line":113,"range":{"start_line":113,"start_character":13,"end_line":113,"end_character":24},"in_reply_to":"3f79a3b5_42b33c0a","updated":"2018-10-23 13:35:26.000000000","message":"yes i was referring to https://github.com/openstack/nova/blob/297de7fb9fbabe74b5305ef0aa82e196d5f48d5e/nova/objects/migrate_data.py#L31","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_2dc04c2b","line":113,"range":{"start_line":113,"start_character":32,"end_line":113,"end_character":52},"in_reply_to":"3f79a3b5_c2e50c09","updated":"2018-10-23 13:35:26.000000000","message":"which is stored in LiveMigrateData\nhttps://github.com/openstack/nova/blob/297de7fb9fbabe74b5305ef0aa82e196d5f48d5e/nova/objects/migrate_data.py#L99\n\nso yes ill update","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_f71ec873","line":114,"updated":"2018-10-21 16:30:37.000000000","message":"Cold migration/resize uses MigrationContext object to consolidate new/old pci requests and allocated devices.\n\nwhat are your thoughts about taking the same approach in live migration by extending the LiveMigrateData object with new/old pci request and allocated devices?\n\nthis will lay the groundwork if/when live-migration will be possible with other (non networking related) PCI devices.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":111,"context_line":"tracker. if claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIF objects in the migrate data object corresponding to the"},{"line_number":114,"context_line":"SR-IOV devices will be updated with the dest host PCI address."},{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call."}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_8d1de0c0","line":114,"in_reply_to":"3f79a3b5_f71ec873","updated":"2018-10-23 13:35:26.000000000","message":"i expclitly do not want to enable or plan to enable pci devices via flavor based pci passthough as part of this spec.\n\nas such i do not want to reuse the coldmigration migration context or any of the move claim logic form cold migration.\n\nmy intent was to sore the newly claimed pci device in \nprofile_json field of the VIFMigrateData\n\nhttps://github.com/openstack/nova/blob/297de7fb9fbabe74b5305ef0aa82e196d5f48d5e/nova/objects/migrate_data.py#L49\n\nwe should not need any modificaitons to the existing data stucures to carry the info we need.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"XML Generation"},{"line_number":120,"context_line":"--------------"},{"line_number":121,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_77b87877","line":118,"updated":"2018-10-21 16:30:37.000000000","message":"Missing a mention on freeing the PCI resources at the source after a successful migration","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":115,"context_line":"If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"XML Generation"},{"line_number":120,"context_line":"--------------"},{"line_number":121,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_4d17c89e","line":118,"in_reply_to":"3f79a3b5_77b87877","updated":"2018-10-23 13:35:26.000000000","message":"yes ill add. good catch.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":145,"context_line":"\t  migration, as the direct mode SR-IOV device will be detachted,"},{"line_number":146,"context_line":"\t  a bond is required in the guest to a transparently live migrateable"},{"line_number":147,"context_line":"\t  interface such as a vsiwtch interface or a indirect mode SR-IOV"},{"line_number":148,"context_line":"\t  device."},{"line_number":149,"context_line":""},{"line_number":150,"context_line":"Alternatives"},{"line_number":151,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_8202b4b8","line":148,"updated":"2018-10-21 16:30:37.000000000","message":"FYI:\nhttps://www.kernel.org/doc/html/latest/networking/net_failover.html","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":145,"context_line":"\t  migration, as the direct mode SR-IOV device will be detachted,"},{"line_number":146,"context_line":"\t  a bond is required in the guest to a transparently live migrateable"},{"line_number":147,"context_line":"\t  interface such as a vsiwtch interface or a indirect mode SR-IOV"},{"line_number":148,"context_line":"\t  device."},{"line_number":149,"context_line":""},{"line_number":150,"context_line":"Alternatives"},{"line_number":151,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_6d7164df","line":148,"in_reply_to":"3f79a3b5_8202b4b8","updated":"2018-10-23 13:35:26.000000000","message":"yes im aware of the net_failover driver\n\nthe note is really only here because of previous question about if this would require bonds.\n\ni think adding support for net_failover would require a new vnic-type in neutron and should be done as a seperate spec. that depens on this one.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"e6899165f7e660e7eb56f071b879645b689d1c16","unresolved":false,"context_lines":[{"line_number":236,"context_line":"  sean-k-mooney"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Other contributors:"},{"line_number":239,"context_line":"  \u003claunchpad-id or None\u003e"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"Work Items"},{"line_number":242,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_4b03540d","line":239,"updated":"2018-10-22 07:32:56.000000000","message":"please add me as a contributor.\nwill reach-out on iRC/Email.\n\nlaunchpad-id : adrian.chiris","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":236,"context_line":"  sean-k-mooney"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Other contributors:"},{"line_number":239,"context_line":"  \u003claunchpad-id or None\u003e"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"Work Items"},{"line_number":242,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_ed46d47e","line":239,"in_reply_to":"3f79a3b5_4b03540d","updated":"2018-10-23 13:35:26.000000000","message":"sure","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"6bfad707eff122f27531720999d4b7e7af0563d3","unresolved":false,"context_lines":[{"line_number":248,"context_line":""},{"line_number":249,"context_line":"This spec has no dependencies but intends to collaborate with the"},{"line_number":250,"context_line":"implementation of NUMA aware live migration [1_]"},{"line_number":251,"context_line":""},{"line_number":252,"context_line":"Testing"},{"line_number":253,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":254,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_22d72025","line":251,"updated":"2018-10-21 16:30:37.000000000","message":"Not sure if this is the correct place to comment but,\nplease note that in order to make macvtap live-migration work, the relevant mechanism driver will need to implement multiple port bindings feature.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"981596b26c05761e0cfff39082ce9f979859c9e0","unresolved":false,"context_lines":[{"line_number":248,"context_line":""},{"line_number":249,"context_line":"This spec has no dependencies but intends to collaborate with the"},{"line_number":250,"context_line":"implementation of NUMA aware live migration [1_]"},{"line_number":251,"context_line":""},{"line_number":252,"context_line":"Testing"},{"line_number":253,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":254,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"3f79a3b5_2d028c0d","line":251,"in_reply_to":"3f79a3b5_22d72025","updated":"2018-10-23 13:35:26.000000000","message":"im not sure if it would be or not.\n\nintially i taught it was due to a db error i was hitting but then after updating my neutron to a slightly newer version that issue went away.\n\nif we do need that it should likely be a seperate neutron blueprint/spec.","commit_id":"4341bf462d30a3291433764eaf87e97311fefd68"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":1,"context_line":".."},{"line_number":2,"context_line":""},{"line_number":3,"context_line":" This work is licensed under a Creative Commons Attribution 3.0 Unported"},{"line_number":4,"context_line":""},{"line_number":5,"context_line":" License."},{"line_number":6,"context_line":""},{"line_number":7,"context_line":" http://creativecommons.org/licenses/by/3.0/legalcode"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Neutron SR-IOV Port LiveMigration"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":16,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a2f657d4","line":13,"range":{"start_line":1,"start_character":0,"end_line":13,"end_character":33},"updated":"2018-10-26 13:08:19.000000000","message":"nit: Can you remove all the extra newlines that have been accidentally added here","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"365587befbdcdcff21c142ae4508e36d75d3c019","unresolved":false,"context_lines":[{"line_number":15,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":16,"context_line":""},{"line_number":17,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":18,"context_line":"out of scope, as a result live migration with SR-IOV while technically"},{"line_number":19,"context_line":"feasible continues to be unsupported. This spec seeks to address this gap"},{"line_number":20,"context_line":"in live migration support for the Libvrit virt driver."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_5870ef09","line":19,"range":{"start_line":18,"start_character":52,"end_line":19,"end_character":9},"updated":"2018-11-05 15:16:46.000000000","message":"nit: , while technically feasible, (commas around the subsentence)","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"365587befbdcdcff21c142ae4508e36d75d3c019","unresolved":false,"context_lines":[{"line_number":17,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":18,"context_line":"out of scope, as a result live migration with SR-IOV while technically"},{"line_number":19,"context_line":"feasible continues to be unsupported. This spec seeks to address this gap"},{"line_number":20,"context_line":"in live migration support for the Libvrit virt driver."},{"line_number":21,"context_line":""},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_5849cfb5","line":20,"range":{"start_line":20,"start_character":34,"end_line":20,"end_character":41},"updated":"2018-11-05 15:16:46.000000000","message":"nit: vrit-\u003evirt","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":31,"context_line":"*     resource claims"},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":34,"context_line":"separately by [1_]"},{"line_number":35,"context_line":""},{"line_number":36,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":37,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_c2e35305","line":34,"range":{"start_line":34,"start_character":14,"end_line":34,"end_character":18},"updated":"2018-10-26 13:08:19.000000000","message":"Use\n\n  [1]_\n\nhttp://rst.ninjs.org/#WzFdXyAKCi4uIFsxXSBUZXN0","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":39,"context_line":""},{"line_number":40,"context_line":""},{"line_number":41,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":42,"context_line":"\t  enabled offloads or Tx/Rx queues that are implemented in hardware"},{"line_number":43,"context_line":"\t  which is not software programmable via the hypervisor e.g. mac"},{"line_number":44,"context_line":"\t  address is not considered hardware state in this context as"},{"line_number":45,"context_line":"\t  libvirt/qemu can set the mac of an SR-IOV device via a standard"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_c21ef30b","line":42,"range":{"start_line":42,"start_character":0,"end_line":42,"end_character":3},"updated":"2018-10-26 13:08:19.000000000","message":"nit: spaces, not tabs","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":46,"context_line":"\t  host level interface."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_e2c76f64","line":49,"range":{"start_line":49,"start_character":37,"end_line":49,"end_character":59},"updated":"2018-10-26 13:08:19.000000000","message":"What other types are there. Might be worth noting what\u0027s considered out of scope here (you don\u0027t need to enumerate them all, but do list some)","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":46,"context_line":"\t  host level interface."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_42b983e9","line":49,"range":{"start_line":49,"start_character":37,"end_line":49,"end_character":38},"updated":"2018-10-26 13:08:19.000000000","message":"nit: two","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"135ccc5ccca1650b70138668620a308f729146b9","unresolved":false,"context_lines":[{"line_number":46,"context_line":"\t  host level interface."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_5f43dc97","line":49,"range":{"start_line":49,"start_character":37,"end_line":49,"end_character":59},"in_reply_to":"3f79a3b5_e2c76f64","updated":"2018-11-06 13:42:40.000000000","message":"i have rephrased this slightly. i wanted to put emphisis on the fact that i am focusing on the two catagories of sriov rather the specifics of each submode.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."},{"line_number":53,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_62bcbfda","line":50,"range":{"start_line":50,"start_character":14,"end_line":50,"end_character":46},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal`` - also s/driect/direct/","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":48,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_82c23b53","line":51,"range":{"start_line":51,"start_character":2,"end_line":51,"end_character":36},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":49,"context_line":"can be done. This spec will focus on 2 categories primarily, direct"},{"line_number":50,"context_line":"passthrough ( vnic_type\u003ddriect|direct-physical) and indirect passthrough"},{"line_number":51,"context_line":"( vnic_type\u003dmacvtap|virtio-forwarder). For simplicity, direct mode and"},{"line_number":52,"context_line":"indirect mode are used instead of vnic_type in the rest of the spec."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":55,"context_line":"performance is achieved at the cost of exposing the guest to the"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a2bdf7d1","line":52,"range":{"start_line":52,"start_character":34,"end_line":52,"end_character":43},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":61,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":62,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":63,"context_line":"virtio-net-PCI device and no hardware state is observed."},{"line_number":64,"context_line":"Indirect mode SR-IOV therefore allows migration of guests."},{"line_number":65,"context_line":""},{"line_number":66,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":67,"context_line":"As mentioned in the introduction it is technically possible to live migrate"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_42874313","line":64,"range":{"start_line":64,"start_character":57,"end_line":64,"end_character":58},"updated":"2018-10-26 13:08:19.000000000","message":"without any workarounds.\n\n(you said in the paragraph above that we\u0027re enabling migration of guests with direct mode and this contradicts that)","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":25733,"name":"Jan Gutter","email":"github@jangutter.com","username":"jangutter"},"change_message_id":"365587befbdcdcff21c142ae4508e36d75d3c019","unresolved":false,"context_lines":[{"line_number":68,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":69,"context_line":"when you do, resources are not correctly claimed."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"When live migrating today migration will fail if the pci mapping is required"},{"line_number":72,"context_line":"to change. Said another way, migration will only succeed if there is a free"},{"line_number":73,"context_line":"pci device on the destination node, with the same pci address as the source"},{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a1546e24","line":71,"updated":"2018-11-05 15:16:46.000000000","message":",","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":73,"context_line":"pci device on the destination node, with the same pci address as the source"},{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of 2 issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and Second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_62b83fc6","line":76,"range":{"start_line":76,"start_character":19,"end_line":76,"end_character":20},"updated":"2018-10-26 13:08:19.000000000","message":"nit: two","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of 2 issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and Second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_82bebbac","line":77,"range":{"start_line":77,"start_character":42,"end_line":77,"end_character":43},"updated":"2018-10-26 13:08:19.000000000","message":"nit: s (now I\u0027m getting into the weeds, I know)","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"135ccc5ccca1650b70138668620a308f729146b9","unresolved":false,"context_lines":[{"line_number":74,"context_line":"node, that is connected to the same physnet and is on the same numa node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of 2 issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and Second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_ba4d5e71","line":77,"range":{"start_line":77,"start_character":42,"end_line":77,"end_character":43},"in_reply_to":"3f79a3b5_82bebbac","updated":"2018-11-06 13:42:40.000000000","message":"no your correct. when i copy past my specs into grammerly it messes up the formating so when i reflow the text somtimes capitalisation gets messed up.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":81,"context_line":"currently unreliable and incorrect when it succeeds."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a2ce5755","line":81,"range":{"start_line":81,"start_character":25,"end_line":81,"end_character":34},"updated":"2018-10-26 13:08:19.000000000","message":"often incorrect?\n\n(theoretically it is possible that all the above - a free PCI device with the same address, physnet, and NUMA affinity - are true, right)","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"135ccc5ccca1650b70138668620a308f729146b9","unresolved":false,"context_lines":[{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":81,"context_line":"currently unreliable and incorrect when it succeeds."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":""},{"line_number":84,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_fae3d638","line":81,"range":{"start_line":81,"start_character":25,"end_line":81,"end_character":34},"in_reply_to":"3f79a3b5_a2ce5755","updated":"2018-11-06 13:42:40.000000000","message":"actually i discovered while openstack can actually move the vm the migration will fail due to a exception that we raise when updating neutron after moving the vm.\n\nso it will still be incorrect as due to how the exception is raise the nova database is never updated to reflect the fact the vm is now running on the destination node.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":85,"context_line":"---------"},{"line_number":86,"context_line":""},{"line_number":87,"context_line":"As a telecom operator with stateful VNF such as a vPE Router"},{"line_number":88,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":89,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":90,"context_line":"flexibility of live migration for maintenance. To that end, as an operator,"},{"line_number":91,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_82ec7bbc","line":88,"range":{"start_line":88,"start_character":16,"end_line":88,"end_character":23},"updated":"2018-10-26 13:08:19.000000000","message":"I\u0027d never encountered this term before. The more you know","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":106,"context_line":"---------------"},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"Building on top of the recently added multiple port binding feature this"},{"line_number":109,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. If claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_4228a3ec","line":109,"range":{"start_line":109,"start_character":37,"end_line":109,"end_character":71},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":109,"context_line":"spec proposes to extend the existing check_can_live_migrate_destination"},{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. If claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIFMigrateData objects in the LiveMigrateData object"},{"line_number":114,"context_line":"corresponding to the SR-IOV devices will be updated with the destination"},{"line_number":115,"context_line":"host PCI address. If the migration should fail after the destination resources"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a23c172d","line":112,"range":{"start_line":112,"start_character":13,"end_line":112,"end_character":47},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. If claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and check_can_live_migrate_destination will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the VIFMigrateData objects in the LiveMigrateData object"},{"line_number":114,"context_line":"corresponding to the SR-IOV devices will be updated with the destination"},{"line_number":115,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_421183ae","line":113,"range":{"start_line":113,"start_character":13,"end_line":113,"end_character":27},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":114,"context_line":"corresponding to the SR-IOV devices will be updated with the destination"},{"line_number":115,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed post_live_migration"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_02070b72","line":117,"range":{"start_line":117,"start_character":0,"end_line":117,"end_character":38},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal``","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":115,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed post_live_migration"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"XML Generation"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_821a3bc4","line":118,"range":{"start_line":118,"start_character":43,"end_line":118,"end_character":62},"updated":"2018-10-26 13:08:19.000000000","message":"nit: ``literal`` assuming this is a function name. If it is, you also need \u0027in\u0027 before it","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"135ccc5ccca1650b70138668620a308f729146b9","unresolved":false,"context_lines":[{"line_number":115,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"rollback_live_migration_at_destination call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed post_live_migration"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"XML Generation"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_d57fcb13","line":118,"range":{"start_line":118,"start_character":43,"end_line":118,"end_character":62},"in_reply_to":"3f79a3b5_821a3bc4","updated":"2018-11-06 13:42:40.000000000","message":"yes its the function that is called on success of the migration.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":144,"context_line":"to disk operation."},{"line_number":145,"context_line":""},{"line_number":146,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":147,"context_line":"          migration, as the direct mode SR-IOV device will be detachted,"},{"line_number":148,"context_line":"          a bond is required in the guest to a transparently live migrateable"},{"line_number":149,"context_line":"          interface such as a vsiwtch interface or a indirect mode SR-IOV"},{"line_number":150,"context_line":"          device. The recently added net_fallback kernel driver is out"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_826f9b27","line":147,"range":{"start_line":147,"start_character":62,"end_line":147,"end_character":71},"updated":"2018-10-26 13:08:19.000000000","message":"detached","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":146,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":147,"context_line":"          migration, as the direct mode SR-IOV device will be detachted,"},{"line_number":148,"context_line":"          a bond is required in the guest to a transparently live migrateable"},{"line_number":149,"context_line":"          interface such as a vsiwtch interface or a indirect mode SR-IOV"},{"line_number":150,"context_line":"          device. The recently added net_fallback kernel driver is out"},{"line_number":151,"context_line":"          of scope of this spec but could also be used."},{"line_number":152,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a26ad714","line":149,"range":{"start_line":149,"start_character":30,"end_line":149,"end_character":37},"updated":"2018-10-26 13:08:19.000000000","message":"vSwitch","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":153,"context_line":"Alternatives"},{"line_number":154,"context_line":"------------"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":157,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":158,"context_line":"  to fall back on cold migration."},{"line_number":159,"context_line":""},{"line_number":160,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":161,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_e24d2fc5","line":158,"range":{"start_line":156,"start_character":0,"end_line":158,"end_character":33},"updated":"2018-10-26 13:08:19.000000000","message":"This is true for every spec and therefore is n4ot really an option. I\u0027d drop this.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"135ccc5ccca1650b70138668620a308f729146b9","unresolved":false,"context_lines":[{"line_number":153,"context_line":"Alternatives"},{"line_number":154,"context_line":"------------"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":157,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":158,"context_line":"  to fall back on cold migration."},{"line_number":159,"context_line":""},{"line_number":160,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":161,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_95c8734d","line":158,"range":{"start_line":156,"start_character":0,"end_line":158,"end_character":33},"in_reply_to":"3f79a3b5_e24d2fc5","updated":"2018-11-06 13:42:40.000000000","message":"all specs can use cold migration good to know :)\ni know doing nothing is always an option i just wanted to record that cold migration was an option instead of live migration even thoug the lifetime of the vms are not the same.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":192,"context_line":"---------------------"},{"line_number":193,"context_line":""},{"line_number":194,"context_line":"Users of direct mode SR-IOV should be aware that auto hotplugging"},{"line_number":195,"context_line":"is not transparent to the guest in exactly the same way that,"},{"line_number":196,"context_line":"suspend is not transparent today."},{"line_number":197,"context_line":""},{"line_number":198,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_8248fbb3","line":195,"range":{"start_line":195,"start_character":60,"end_line":195,"end_character":61},"updated":"2018-10-26 13:08:19.000000000","message":"nit: drop","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":199,"context_line":"Performance Impact"},{"line_number":200,"context_line":"------------------"},{"line_number":201,"context_line":""},{"line_number":202,"context_line":"This should not significantly impact the performance of a live migration"},{"line_number":203,"context_line":"a minor overhead will be incurred in claiming the resource and updating the xml"},{"line_number":204,"context_line":""},{"line_number":205,"context_line":"Other deployer impact"},{"line_number":206,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_a243b796","line":203,"range":{"start_line":202,"start_character":72,"end_line":203,"end_character":1},"updated":"2018-10-26 13:08:19.000000000","message":". A","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":205,"context_line":"Other deployer impact"},{"line_number":206,"context_line":"---------------------"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"if a new flavour extra spec or image property is added to enable this"},{"line_number":209,"context_line":"functionality then a deployer will have to update there flavours/images"},{"line_number":210,"context_line":"to use this feature."},{"line_number":211,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_62c7df12","line":208,"range":{"start_line":208,"start_character":0,"end_line":208,"end_character":2},"updated":"2018-10-26 13:08:19.000000000","message":"nit: If","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":209,"context_line":"functionality then a deployer will have to update there flavours/images"},{"line_number":210,"context_line":"to use this feature."},{"line_number":211,"context_line":""},{"line_number":212,"context_line":"similarly, if a config option is added instead they will have to modify"},{"line_number":213,"context_line":"their config."},{"line_number":214,"context_line":""},{"line_number":215,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_02caab17","line":212,"range":{"start_line":212,"start_character":0,"end_line":212,"end_character":1},"updated":"2018-10-26 13:08:19.000000000","message":"nit: S","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"065dd3bf1851d241ef813303a352bec799803ad9","unresolved":false,"context_lines":[{"line_number":262,"context_line":"This feature will be tested primarily via unit and functional test."},{"line_number":263,"context_line":"as SR-IOV testing is not available in the gate tempest test will not"},{"line_number":264,"context_line":"be possible. Third party ci could be implemented but that is not part"},{"line_number":265,"context_line":"of the scope of this spec."},{"line_number":266,"context_line":""},{"line_number":267,"context_line":"Documentation Impact"},{"line_number":268,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":5,"id":"3f79a3b5_02e38ba1","line":265,"updated":"2018-10-26 13:08:19.000000000","message":"You should mention your investigation into that SR-IOV network driver that was added recently.","commit_id":"1828f3ac909220696bd60388e05463121642994b"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":4,"context_line":" http://creativecommons.org/licenses/by/3.0/legalcode"},{"line_number":5,"context_line":""},{"line_number":6,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":7,"context_line":"Neutron SR-IOV Port LiveMigration"},{"line_number":8,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_0686fcb8","line":7,"range":{"start_line":7,"start_character":20,"end_line":7,"end_character":33},"updated":"2018-11-29 17:42:04.000000000","message":"nit: live migration","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":13,"context_line":"out of scope, as a result live migration with SR-IOV, while technically"},{"line_number":14,"context_line":"feasible, continues to be unsupported. This spec seeks to address this gap"},{"line_number":15,"context_line":"in live migration support for the Libvirt virt driver."},{"line_number":16,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_064d9c92","line":13,"range":{"start_line":12,"start_character":57,"end_line":13,"end_character":12},"updated":"2018-11-29 17:42:04.000000000","message":"do we have any spec telling this ? I don\u0027t remember this consensus and why.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":13,"context_line":"out of scope, as a result live migration with SR-IOV, while technically"},{"line_number":14,"context_line":"feasible, continues to be unsupported. This spec seeks to address this gap"},{"line_number":15,"context_line":"in live migration support for the Libvirt virt driver."},{"line_number":16,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_a642a869","line":13,"range":{"start_line":13,"start_character":12,"end_line":13,"end_character":15},"updated":"2018-11-29 17:42:04.000000000","message":"nit: \". As\"","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":13,"context_line":"out of scope, as a result live migration with SR-IOV, while technically"},{"line_number":14,"context_line":"feasible, continues to be unsupported. This spec seeks to address this gap"},{"line_number":15,"context_line":"in live migration support for the Libvirt virt driver."},{"line_number":16,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_6f52236d","line":13,"range":{"start_line":12,"start_character":57,"end_line":13,"end_character":12},"in_reply_to":"3f79a3b5_064d9c92","updated":"2018-11-29 19:55:30.000000000","message":"the spec that intoduced this feature was \n\nhttps://specs.openstack.org/openstack/nova-specs/specs/juno/implemented/pci-passthrough-sriov.html\n\ni does say that macvtap mode would enable migration but\nthe spec never addressed any of the modification required\nand the claims logoic required was not implemented.\n\nfor driect mode sriov it was not specified but implied that it did not support migration.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":13,"context_line":"out of scope, as a result live migration with SR-IOV, while technically"},{"line_number":14,"context_line":"feasible, continues to be unsupported. This spec seeks to address this gap"},{"line_number":15,"context_line":"in live migration support for the Libvirt virt driver."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":""},{"line_number":18,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_261bf885","line":15,"range":{"start_line":14,"start_character":39,"end_line":15,"end_character":17},"updated":"2018-11-29 17:42:04.000000000","message":"if we said \"out of scope\", why should it be now in scope ? \n\nI mean, I understand your point, I\u0027m just trying to explain that we shouldn\u0027t tell that the live migration was out of scope, just that we didn\u0027t have time for it before.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV, live migration was declared"},{"line_number":13,"context_line":"out of scope, as a result live migration with SR-IOV, while technically"},{"line_number":14,"context_line":"feasible, continues to be unsupported. This spec seeks to address this gap"},{"line_number":15,"context_line":"in live migration support for the Libvirt virt driver."},{"line_number":16,"context_line":""},{"line_number":17,"context_line":""},{"line_number":18,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_4f4967b7","line":15,"range":{"start_line":14,"start_character":39,"end_line":15,"end_character":17},"in_reply_to":"3f79a3b5_261bf885","updated":"2018-11-29 19:55:30.000000000","message":"ok i think it really was just a time thing\nthe origianl intent was to support live migration so ill rephases this and the intro to refelect that.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":26,"context_line":"*     resource claims"},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":29,"context_line":"separately by [1]_"},{"line_number":30,"context_line":""},{"line_number":31,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":32,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_a91efb5e","line":29,"range":{"start_line":29,"start_character":14,"end_line":29,"end_character":18},"updated":"2018-11-29 17:42:04.000000000","message":"in general, we try to provide an explicit markup block and just not a footnote, but okay.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":30,"context_line":""},{"line_number":31,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":32,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":33,"context_line":"state."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":""},{"line_number":36,"context_line":".. note:: hardware state in this context refers to any nic state such as"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_29374bdf","line":33,"updated":"2018-11-29 17:42:04.000000000","message":"it\u0027s an hypervisor feature, right ? Just tell it if so.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":30,"context_line":""},{"line_number":31,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":32,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":33,"context_line":"state."},{"line_number":34,"context_line":""},{"line_number":35,"context_line":""},{"line_number":36,"context_line":".. note:: hardware state in this context refers to any nic state such as"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_4ff7275d","line":33,"in_reply_to":"3f79a3b5_29374bdf","updated":"2018-11-29 19:55:30.000000000","message":"not really its more of a technology limitation.\n\n qemu and other hypervisors would need to know the inner working of each device to be able to transfer the state.\n\nso it would require explicit hardware,firmware and hypervior support to copy the hardware state.\n\nthe PCI sig did not stanise such a feature when defining sriov but newer technoligys such as vfio-mediated devices have been developed to enable this in the future.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":29577,"name":"Parav Pandit","email":"parav@mellanox.com"},"change_message_id":"b5e2fd0c7d668c86177dad9fad0ba70dc1f4fb21","unresolved":false,"context_lines":[{"line_number":42,"context_line":""},{"line_number":43,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":44,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"},{"line_number":45,"context_line":"direct passthrough (``vnic_type\u003ddriect|direct-physical``) and indirect"},{"line_number":46,"context_line":"passthrough (``vnic_type\u003dmacvtap|virtio-forwarder``). For simplicity, direct"},{"line_number":47,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":48,"context_line":"of the spec."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_8c0ef42d","line":45,"updated":"2018-11-26 23:06:30.000000000","message":"vnic_type\u003ddirect|direct-physical","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":45,"context_line":"direct passthrough (``vnic_type\u003ddriect|direct-physical``) and indirect"},{"line_number":46,"context_line":"passthrough (``vnic_type\u003dmacvtap|virtio-forwarder``). For simplicity, direct"},{"line_number":47,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":48,"context_line":"of the spec."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":51,"context_line":"performance is achieved at the cost of exposing the guest to the"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_89ee9f1e","line":48,"updated":"2018-11-29 17:42:04.000000000","message":"This assumes that SR-IOV live migration will only be supported for those modes, besides the fact that only NIC devices will be supported.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":45,"context_line":"direct passthrough (``vnic_type\u003ddriect|direct-physical``) and indirect"},{"line_number":46,"context_line":"passthrough (``vnic_type\u003dmacvtap|virtio-forwarder``). For simplicity, direct"},{"line_number":47,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":48,"context_line":"of the spec."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":51,"context_line":"performance is achieved at the cost of exposing the guest to the"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_cf5ed74f","line":48,"in_reply_to":"3f79a3b5_89ee9f1e","updated":"2018-11-29 19:55:30.000000000","message":"yes the scope of this spec is limited to enableing migratin of sriov nic via neutron.\n\nthe same techniques could be applied to other sriov device but i wanted to keep that out of scope of this spec and just support the neutron case this cycle.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":29577,"name":"Parav Pandit","email":"parav@mellanox.com"},"change_message_id":"b5e2fd0c7d668c86177dad9fad0ba70dc1f4fb21","unresolved":false,"context_lines":[{"line_number":47,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":48,"context_line":"of the spec."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":51,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":52,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":53,"context_line":"state direct mode SR-IOV cannot be conventionally live migrated."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_4c4bfc5c","line":50,"updated":"2018-11-26 23:06:30.000000000","message":"mode SR-IOV,","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":64,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":65,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":66,"context_line":"device an exception is raised after the vm has been sucessfully migrated to the"},{"line_number":67,"context_line":"destination in the post migration cleanup."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":70,"context_line":"required to change. Said another way, migration will only succeed if there is"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_893c5f92","line":67,"updated":"2018-11-29 17:42:04.000000000","message":"I\u0027m unclear, you mention the Nova tech debt where we don\u0027t claim SR-IOV resources on live-migration ?\n\nIsn\u0027t the whole purpose of the spec to address this gap and the rest of factors is just the context ?","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":64,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":65,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":66,"context_line":"device an exception is raised after the vm has been sucessfully migrated to the"},{"line_number":67,"context_line":"destination in the post migration cleanup."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":70,"context_line":"required to change. Said another way, migration will only succeed if there is"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_4f188786","line":67,"in_reply_to":"3f79a3b5_893c5f92","updated":"2018-11-29 19:55:30.000000000","message":"yes it is. \n\ncurrently if you have a vm with a macvtap sriov interface and you ask openstack to migrate it it will move the vm and then we fail after the vm is running on the destination node because we check if we claimed the pci device before updating the port binding in neutron but after migrating the vm which result in an exception that causes the migration to fail in post migrate.\n\nso the fact is this feature is already 90% implemented.\nthe only thing that is missing is the claiming of the resources and the handeling of the different sucess/error cases.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":77,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":80,"context_line":"currently unsupported and incorrect even when the vm is successfully"},{"line_number":81,"context_line":"moved."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_09484f29","line":80,"range":{"start_line":80,"start_character":10,"end_line":80,"end_character":21},"updated":"2018-11-29 17:42:04.000000000","message":"we technically allow live-migrating instances, right ? So, it\u0027s not about \u0027unsupporting\u0027 things, rather fixing a wrong behaviour.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":77,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":80,"context_line":"currently unsupported and incorrect even when the vm is successfully"},{"line_number":81,"context_line":"moved."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_fa0fcf29","line":80,"range":{"start_line":80,"start_character":10,"end_line":80,"end_character":21},"in_reply_to":"3f79a3b5_09484f29","updated":"2018-11-29 19:55:30.000000000","message":"correct we do allow the migration but it will away fail in a way that is non obvious.\n\nthe vm will be running on the if it landed on a host with a free pci devicen. but nova will think it is runing on the souce node.\n\ninstance operation on the vm at this point are undefined\nso it will require manual admin intervention to fix.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"0c5051e53637a6e05497752473ed3e02b213bf77","unresolved":false,"context_lines":[{"line_number":77,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"As a result of the above issues, SR-IOV live migration in OpenStack is"},{"line_number":80,"context_line":"currently unsupported and incorrect even when the vm is successfully"},{"line_number":81,"context_line":"moved."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"dfd5e7cf_2ae7f8e6","line":80,"range":{"start_line":80,"start_character":10,"end_line":80,"end_character":21},"in_reply_to":"3f79a3b5_fa0fcf29","updated":"2019-01-10 14:59:28.000000000","message":"To be clear, I think \"unsupported\" is the correct term here. Users may try it and it could work, but if they came to us with complaints we\u0027d surely tell them \"don\u0027t do that\".\n\nThat said, we should call out that this feature might break the rare user that was relying on this edgecase. I\u0027ve called that out in the code review.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"79bd12c37ea44c562789055dc1d38d7e2d0ef8c8","unresolved":false,"context_lines":[{"line_number":106,"context_line":"---------------"},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"Building on top of the recently added multiple port binding feature this"},{"line_number":109,"context_line":"spec proposes to extend the existing ``check_can_live_migrate_destination``"},{"line_number":110,"context_line":"function to claim SR-IOV devices on the destination node via the PCI resource"},{"line_number":111,"context_line":"tracker. If claiming fails then the partially claimed resources will be"},{"line_number":112,"context_line":"released and ``check_can_live_migrate_destination`` will fail. If the claiming"},{"line_number":113,"context_line":"succeeds the ``VIFMigrateData`` objects in the ``LiveMigrateData`` object"},{"line_number":114,"context_line":"corresponding to the SR-IOV devices will be updated with the destination"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_6b2133b2","line":111,"range":{"start_line":109,"start_character":17,"end_line":111,"end_character":8},"updated":"2018-11-22 12:39:25.000000000","message":"As far as i see it we have two options to perform the claim here.\n\nOn a per VIF basis:\n1. use the same PCI request as the one which was used to claim the VIF on the source when the instance was created\n\n2. create a new PCI request (with a new UUID)\n\nthe latter will require data model changes as we will need to track the new and old  pci requests to allow updating instance_extra DB with the new information.\n\ndownside with the former is that we will have a point in time where the same PCI request is allocated on host A and claimed on Host B.\nSelf Healing in case of unexpected error is done via update_available_resources() in resource tracker","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"ad318f7f5afbba6634f1b128bb849e170d9a5ac8","unresolved":false,"context_lines":[{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Indirect Mode"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_64ed1d22","line":119,"updated":"2018-12-04 09:19:16.000000000","message":"So, in the POC code:\nduring check_can_live_migrate_destination() we claim the PCI resources if needed and update VIFMigrateData.\n\nin case of a successful migration two things happen in regards to the PCI resources:\n1. PCI resources are freed on the source during post_live_migration()\n2. PCI resources are allocated (PCI resource state change claimed -\u003e allocated) on the destination in post_live_migration_at_destination()\n\nin case of a failed migration we free the claimed PCI resources on the destination in rollback_live_migration_at_destination().\n\nminimizing the dependency on the resource_tracker periodic job: update_available_resource() to clean/sort things for us.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"7805c4ac2c9275c2ffcec826643a2e34e1dd6403","unresolved":false,"context_lines":[{"line_number":116,"context_line":"have been claimed they must be released in the"},{"line_number":117,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Indirect Mode"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_db9639ef","line":119,"in_reply_to":"3f79a3b5_64ed1d22","updated":"2018-12-04 23:13:47.000000000","message":"thanks i will intergete that into the next version.\nthat is on my todo list for tomorow to have it ready before the next nova team meeting.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":117,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Indirect Mode"},{"line_number":123,"context_line":"-------------"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_09612fa3","line":120,"updated":"2018-11-29 17:42:04.000000000","message":"please note there would be some rolling upgrade issue where some old source node would get some information it couldn\u0027t handle. How do you plan to address this ?","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":117,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":118,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":119,"context_line":"(clean up source)."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":""},{"line_number":122,"context_line":"Indirect Mode"},{"line_number":123,"context_line":"-------------"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_1a608bd1","line":120,"in_reply_to":"3f79a3b5_09612fa3","updated":"2018-11-29 19:55:30.000000000","message":"that should not happen as we plan to expliticly check in the conductor that both node support this new workflow the same way we do form multiple port bindings.\n\nhttps://github.com/openstack/nova/blob/master/nova/conductor/tasks/live_migrate.py#L245-L249\n\nso we will only include the info if both the source and dest nodes support it.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":29577,"name":"Parav Pandit","email":"parav@mellanox.com"},"change_message_id":"b5e2fd0c7d668c86177dad9fad0ba70dc1f4fb21","unresolved":false,"context_lines":[{"line_number":122,"context_line":"Indirect Mode"},{"line_number":123,"context_line":"-------------"},{"line_number":124,"context_line":""},{"line_number":125,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":126,"context_line":""},{"line_number":127,"context_line":""},{"line_number":128,"context_line":"Direct Mode"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_2cc24003","line":125,"updated":"2018-11-26 23:06:30.000000000","message":"SR-IOV.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":161,"context_line":""},{"line_number":162,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":163,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":164,"context_line":"  to fall back on cold migration."},{"line_number":165,"context_line":""},{"line_number":166,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":167,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_c90017b4","line":164,"updated":"2018-11-29 17:42:04.000000000","message":"that\u0027s not really an alternative ;)","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":161,"context_line":""},{"line_number":162,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":163,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":164,"context_line":"  to fall back on cold migration."},{"line_number":165,"context_line":""},{"line_number":166,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":167,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_daae93e4","line":164,"in_reply_to":"3f79a3b5_c90017b4","updated":"2018-11-29 19:55:30.000000000","message":"well its another way to move the vm :)\nbut its not an alternitive way of doing live migration with sriov. \ni can remove it if you like.","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"61e2f25acf0272087974c551d672bf2c27e757c3","unresolved":false,"context_lines":[{"line_number":233,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"},{"line_number":234,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":235,"context_line":"however, as that will require both the source and dest node to"},{"line_number":236,"context_line":"this feature, it will have no impact for this release."},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Implementation"},{"line_number":239,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_691ea310","line":236,"updated":"2018-11-29 17:42:04.000000000","message":"how do you plan to allow this ?","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"049daf04f6d8e0844b7c87adce2473b565fc085e","unresolved":false,"context_lines":[{"line_number":233,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"},{"line_number":234,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":235,"context_line":"however, as that will require both the source and dest node to"},{"line_number":236,"context_line":"this feature, it will have no impact for this release."},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Implementation"},{"line_number":239,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3f79a3b5_9adfbb6c","line":236,"in_reply_to":"3f79a3b5_691ea310","updated":"2018-11-29 19:55:30.000000000","message":"i was trying to say once both nodes are upgraded to stien you will be able to use this to simply future upgrades\nbut it will have no impact for stien as its currently broken and this wont make it worse as we will disable this logic and use the old workfolw if both nodes dont support it.\n\ne.g. the same thing we do for multiple port bindigns \nhttps://github.com/openstack/nova/blob/master/nova/conductor/tasks/live_migrate.py#L245-L249","commit_id":"d47af92879a1d9d62ac2b2d7071c9ffd557ddcc0"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV by [0]_, live migration was not"},{"line_number":13,"context_line":"directly addressed as part of the proposed changes."},{"line_number":14,"context_line":"As a result while live migration with SR-IOV is technically feasible,"},{"line_number":15,"context_line":"it remains unsupported by the libvirt driver. This spec seeks to address"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f662bca7","line":12,"range":{"start_line":12,"start_character":46,"end_line":12,"end_character":47},"updated":"2018-12-10 18:19:39.000000000","message":"]","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV by [0]_, live migration was not"},{"line_number":13,"context_line":"directly addressed as part of the proposed changes."},{"line_number":14,"context_line":"As a result while live migration with SR-IOV is technically feasible,"},{"line_number":15,"context_line":"it remains unsupported by the libvirt driver. This spec seeks to address"},{"line_number":16,"context_line":"this gap in live migration support for the Libvirt virt driver."}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d62a202c","line":13,"updated":"2018-12-10 18:19:39.000000000","message":"Weird line break - either make it a proper second para, or remove the newline :)","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":13,"context_line":"directly addressed as part of the proposed changes."},{"line_number":14,"context_line":"As a result while live migration with SR-IOV is technically feasible,"},{"line_number":15,"context_line":"it remains unsupported by the libvirt driver. This spec seeks to address"},{"line_number":16,"context_line":"this gap in live migration support for the Libvirt virt driver."},{"line_number":17,"context_line":""},{"line_number":18,"context_line":""},{"line_number":19,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_56edb0fe","line":16,"range":{"start_line":16,"start_character":43,"end_line":16,"end_character":50},"updated":"2018-12-10 18:19:39.000000000","message":"libvirt","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":16,"context_line":"this gap in live migration support for the Libvirt virt driver."},{"line_number":17,"context_line":""},{"line_number":18,"context_line":""},{"line_number":19,"context_line":"Problem description"},{"line_number":20,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"Live Migration with SR-IOV devices has several complicating factors."}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b68e84d4","line":19,"updated":"2018-12-10 18:19:39.000000000","message":"I think I kinda want this section split into three: an introduction that explains direct vs indirect, one section for direct mode and one section for indirect, and each can explain the current state of things for that kind of live migration.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":26,"context_line":"*     SR-IOV mode"},{"line_number":27,"context_line":"*     resource claims"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":30,"context_line":"separately by [1]_"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_2a4f6e8d","line":29,"range":{"start_line":29,"start_character":0,"end_line":29,"end_character":4},"updated":"2018-12-10 16:00:48.000000000","message":"NUMA","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":29,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":30,"context_line":"separately by [1]_"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":33,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":34,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":35,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":""},{"line_number":38,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":39,"context_line":"          enabled offloads or Tx/Rx queues that are implemented in hardware"},{"line_number":40,"context_line":"          which is not software programmable via the hypervisor e.g. mac"},{"line_number":41,"context_line":"          address is not considered hardware state in this context as"},{"line_number":42,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":43,"context_line":"          host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_140d1eda","line":43,"range":{"start_line":32,"start_character":0,"end_line":43,"end_character":31},"updated":"2018-12-10 18:19:39.000000000","message":"I\u0027d invert those two paras - first explain what hardware stat e is, then explain why we can\u0027t handle it. And I\u0027d drop the note as well.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":29,"context_line":"numa affinity is out of the scope of this spec and will be addressed"},{"line_number":30,"context_line":"separately by [1]_"},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":33,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":34,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":35,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":""},{"line_number":38,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":39,"context_line":"          enabled offloads or Tx/Rx queues that are implemented in hardware"},{"line_number":40,"context_line":"          which is not software programmable via the hypervisor e.g. mac"},{"line_number":41,"context_line":"          address is not considered hardware state in this context as"},{"line_number":42,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":43,"context_line":"          host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_7afb4648","line":43,"range":{"start_line":32,"start_character":0,"end_line":43,"end_character":31},"in_reply_to":"3f79a3b5_140d1eda","updated":"2018-12-20 17:42:22.000000000","message":"i put the definition of hardware state as a note because, what hardware state is, is irrelevant to the over all problem statement. I wanted to provide context but did not want to distract by going into detail about hardware state.\n\ni will however move this later in the spec.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":42,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":43,"context_line":"          host level interface."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"The SR-IOV mode of a neutron port directly impacts the how a live migration"},{"line_number":46,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"},{"line_number":47,"context_line":"direct passthrough (``vnic_type\u003ddirect|direct-physical``) and indirect"},{"line_number":48,"context_line":"passthrough (``vnic_type\u003dmacvtap|virtio-forwarder``). For simplicity, direct"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_aa305e06","line":45,"range":{"start_line":45,"start_character":51,"end_line":45,"end_character":54},"updated":"2018-12-10 16:00:48.000000000","message":"drop","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":52,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":53,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":54,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":55,"context_line":"state direct mode SR-IOV, cannot be conventionally live migrated."},{"line_number":56,"context_line":"This spec will provide a workaround to enable this configuration."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_8a02420c","line":55,"range":{"start_line":55,"start_character":0,"end_line":55,"end_character":32},"updated":"2018-12-10 16:00:48.000000000","message":"state, direct mode SR-IOV cannot be\n\n(move the comma or this reads wacky)","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":52,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":53,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":54,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":55,"context_line":"state direct mode SR-IOV, cannot be conventionally live migrated."},{"line_number":56,"context_line":"This spec will provide a workaround to enable this configuration."},{"line_number":57,"context_line":""},{"line_number":58,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_36a834d8","line":55,"range":{"start_line":55,"start_character":0,"end_line":55,"end_character":32},"in_reply_to":"3f79a3b5_8a02420c","updated":"2018-12-10 18:19:39.000000000","message":"Oooh, I had 2 nits here, but this explains what Sean meant.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":65,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":66,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":67,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":68,"context_line":"device an exception is raised after the vm has been sucessfully migrated to the"},{"line_number":69,"context_line":"destination in the post migration cleanup."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"When live migrating today, migration will also fail if the pci mapping is"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_9676e890","line":68,"range":{"start_line":68,"start_character":40,"end_line":68,"end_character":42},"updated":"2018-12-10 18:19:39.000000000","message":"VM","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":68,"context_line":"device an exception is raised after the vm has been sucessfully migrated to the"},{"line_number":69,"context_line":"destination in the post migration cleanup."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":72,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":73,"context_line":"a free pci device on the destination node, with the same pci address as the"},{"line_number":74,"context_line":"source node, that is connected to the same physnet and is on the same numa"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_568cd07f","line":71,"range":{"start_line":71,"start_character":59,"end_line":71,"end_character":62},"updated":"2018-12-10 18:19:39.000000000","message":"PCI","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":70,"context_line":""},{"line_number":71,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":72,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":73,"context_line":"a free pci device on the destination node, with the same pci address as the"},{"line_number":74,"context_line":"source node, that is connected to the same physnet and is on the same numa"},{"line_number":75,"context_line":"node."},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_1682d84b","line":73,"range":{"start_line":73,"start_character":7,"end_line":73,"end_character":10},"updated":"2018-12-10 18:19:39.000000000","message":"PCI","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":70,"context_line":""},{"line_number":71,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":72,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":73,"context_line":"a free pci device on the destination node, with the same pci address as the"},{"line_number":74,"context_line":"source node, that is connected to the same physnet and is on the same numa"},{"line_number":75,"context_line":"node."},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d6e5403e","line":73,"range":{"start_line":73,"start_character":57,"end_line":73,"end_character":60},"updated":"2018-12-10 18:19:39.000000000","message":"PCI","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":71,"context_line":"When live migrating today, migration will also fail if the pci mapping is"},{"line_number":72,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":73,"context_line":"a free pci device on the destination node, with the same pci address as the"},{"line_number":74,"context_line":"source node, that is connected to the same physnet and is on the same numa"},{"line_number":75,"context_line":"node."},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"This is because of two issues. First nova does not correctly claim the"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_76e8ec12","line":74,"range":{"start_line":74,"start_character":70,"end_line":74,"end_character":74},"updated":"2018-12-10 18:19:39.000000000","message":"NUMA","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":79,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"As a result of the above issues, SR-IOV live migration in the libvirt driver"},{"line_number":82,"context_line":"is currently incomplete and incorrect even when the vm is successfully"},{"line_number":83,"context_line":"moved."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_56d13059","line":82,"range":{"start_line":82,"start_character":52,"end_line":82,"end_character":54},"updated":"2018-12-10 18:19:39.000000000","message":"VM","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":81,"context_line":"As a result of the above issues, SR-IOV live migration in the libvirt driver"},{"line_number":82,"context_line":"is currently incomplete and incorrect even when the vm is successfully"},{"line_number":83,"context_line":"moved."},{"line_number":84,"context_line":""},{"line_number":85,"context_line":""},{"line_number":86,"context_line":"Use Cases"},{"line_number":87,"context_line":"---------"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b0468c55","line":84,"updated":"2018-12-10 16:00:48.000000000","message":"++ good evaluation of the current issues here. I would personally like to see reference to the live migration w/ CPU pinning/NUMA issue, since they are very similar and would help provide a more holistic view, but that\u0027s a nice-to-have","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":97,"context_line":"As the provider of a cloud with a hardware offloaded vSwitch that leverages"},{"line_number":98,"context_line":"indirect mode SR-IOV, I want to offer the performance it enables to my"},{"line_number":99,"context_line":"customers but also desire the flexibility to be able to transparently migrate"},{"line_number":100,"context_line":"guests without disrupting traffic to enable maintenance."},{"line_number":101,"context_line":""},{"line_number":102,"context_line":"Proposed change"},{"line_number":103,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f0588426","line":100,"range":{"start_line":100,"start_character":0,"end_line":100,"end_character":56},"updated":"2018-12-10 16:00:48.000000000","message":"I was going to note that this isn\u0027t true, then I re-read and spotted that this is in reference to indirect mode. nvm","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"This spec proposes addressing the problem statement in several steps."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Resource claims"},{"line_number":108,"context_line":"---------------"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Building on top of the recently added multiple port binding feature this"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b6d684f7","line":107,"updated":"2018-12-10 18:19:39.000000000","message":"The way you wrote this section makes it look like it only applies to Direct mode, but there there\u0027s another Direct Mode section below. This is ignorance on my part, but in indirect mode the macvtap device is still connected (plugged? What\u0027s the right word?) to a VF on the host? Does neutron handle that plugging for us? Why don\u0027t we need to claim the underlying VF on the dest? I\u0027d rework the Proposed Section a bit to make it more explicit what\u0027s needed for direct and indirect. I\u0027d put Direct and Indirect as the only two subsections, and then sub-susbsections in each of those describing resrouces claims, XML generation, and anything else that needs to be done for each. Idealler the sub-subsections would be identical, and if one doesn\u0027t apply for a particular mode, have a quick sentence about why that\u0027s the case.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"This spec proposes addressing the problem statement in several steps."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Resource claims"},{"line_number":108,"context_line":"---------------"},{"line_number":109,"context_line":""},{"line_number":110,"context_line":"Building on top of the recently added multiple port binding feature this"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d5eb4b0d","line":107,"in_reply_to":"3f79a3b5_b6d684f7","updated":"2018-12-20 17:42:22.000000000","message":"there are several qustions here.\n\nThe macvtap is created form the netdev associated with the VF.The macvtap is created by libvirt using the xml generated by nova.\n\nwe do need to claim the the vf in both direct and indirect mode. the resource claims section is identical for both modes which is why i have intentionally not split this into two sections as you suggested.\n\nin the macvtap case you just don need to state the pci address of the vf in the xml. for macvtap mode we specify the name of the vf netdev instead of the pci address.\n\nim not sure what makes you think this apply only to direct mode as i intentioally didn not specify an sriov mode in the resouce claims section. the indirect mode and direct mode sections at line 127 and 133 are not sub headings of resouce claims.\n\ni will make that more clear.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":117,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":118,"context_line":"have been claimed they must be released in the"},{"line_number":119,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":120,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":121,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":122,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":123,"context_line":"success and failure case we do not need to rely on the"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_56ba900b","line":120,"range":{"start_line":120,"start_character":16,"end_line":120,"end_character":21},"updated":"2018-12-10 18:19:39.000000000","message":"SR-IOV","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":121,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":122,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":123,"context_line":"success and failure case we do not need to rely on the"},{"line_number":124,"context_line":"update_available_resource periodic task to heal the allocations/claims."},{"line_number":125,"context_line":""},{"line_number":126,"context_line":""},{"line_number":127,"context_line":"Indirect Mode"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f6b4fc1f","line":124,"range":{"start_line":124,"start_character":0,"end_line":124,"end_character":25},"updated":"2018-12-10 18:19:39.000000000","message":"``update_available_resource``","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":134,"context_line":"-------------"},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"For direct mode SR-IOV, to enable live migration the SR-IOV devices must"},{"line_number":137,"context_line":"be first detached on the source after pre_live_migrate(setup_dest)"},{"line_number":138,"context_line":"and then reattached in post_live_migration_at_destination (finish dest)."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"This mimics the existing suspend [2_] and resume [3_] workflow where"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_3644b419","line":137,"range":{"start_line":137,"start_character":38,"end_line":137,"end_character":66},"updated":"2018-12-10 18:19:39.000000000","message":"``pre_live_migration``","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":135,"context_line":""},{"line_number":136,"context_line":"For direct mode SR-IOV, to enable live migration the SR-IOV devices must"},{"line_number":137,"context_line":"be first detached on the source after pre_live_migrate(setup_dest)"},{"line_number":138,"context_line":"and then reattached in post_live_migration_at_destination (finish dest)."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"This mimics the existing suspend [2_] and resume [3_] workflow where"},{"line_number":141,"context_line":"by we workaround QEMUs inability to save device state during a suspend"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f6495c2f","line":138,"range":{"start_line":138,"start_character":58,"end_line":138,"end_character":71},"updated":"2018-12-10 18:19:39.000000000","message":"I\u0027d just drop this.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":135,"context_line":""},{"line_number":136,"context_line":"For direct mode SR-IOV, to enable live migration the SR-IOV devices must"},{"line_number":137,"context_line":"be first detached on the source after pre_live_migrate(setup_dest)"},{"line_number":138,"context_line":"and then reattached in post_live_migration_at_destination (finish dest)."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"This mimics the existing suspend [2_] and resume [3_] workflow where"},{"line_number":141,"context_line":"by we workaround QEMUs inability to save device state during a suspend"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_563ff082","line":138,"range":{"start_line":138,"start_character":23,"end_line":138,"end_character":57},"updated":"2018-12-10 18:19:39.000000000","message":"``post_live_migration_at_destination``","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":137,"context_line":"be first detached on the source after pre_live_migrate(setup_dest)"},{"line_number":138,"context_line":"and then reattached in post_live_migration_at_destination (finish dest)."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"This mimics the existing suspend [2_] and resume [3_] workflow where"},{"line_number":141,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":142,"context_line":"to disk operation."},{"line_number":143,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b659a400","line":140,"range":{"start_line":140,"start_character":33,"end_line":140,"end_character":37},"updated":"2018-12-10 18:19:39.000000000","message":"[2]_","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":137,"context_line":"be first detached on the source after pre_live_migrate(setup_dest)"},{"line_number":138,"context_line":"and then reattached in post_live_migration_at_destination (finish dest)."},{"line_number":139,"context_line":""},{"line_number":140,"context_line":"This mimics the existing suspend [2_] and resume [3_] workflow where"},{"line_number":141,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":142,"context_line":"to disk operation."},{"line_number":143,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d65c60f0","line":140,"range":{"start_line":140,"start_character":49,"end_line":140,"end_character":53},"updated":"2018-12-10 18:19:39.000000000","message":"[3]_","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":143,"context_line":""},{"line_number":144,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":145,"context_line":"          migration, as the direct mode SR-IOV device will be detached,"},{"line_number":146,"context_line":"          a bond is required in the guest to a transparently live migrateable"},{"line_number":147,"context_line":"          interface such as a vSwitch interface or a indirect mode SR-IOV"},{"line_number":148,"context_line":"          device. The recently added net_fallback kernel driver is out"},{"line_number":149,"context_line":"          of scope of this spec but could also be used."}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_9076f090","line":146,"range":{"start_line":146,"start_character":66,"end_line":146,"end_character":77},"updated":"2018-12-10 16:00:48.000000000","message":"migratable","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":145,"context_line":"          migration, as the direct mode SR-IOV device will be detached,"},{"line_number":146,"context_line":"          a bond is required in the guest to a transparently live migrateable"},{"line_number":147,"context_line":"          interface such as a vSwitch interface or a indirect mode SR-IOV"},{"line_number":148,"context_line":"          device. The recently added net_fallback kernel driver is out"},{"line_number":149,"context_line":"          of scope of this spec but could also be used."},{"line_number":150,"context_line":""},{"line_number":151,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f0864460","line":148,"range":{"start_line":148,"start_character":37,"end_line":148,"end_character":49},"updated":"2018-12-10 16:00:48.000000000","message":"link","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":152,"context_line":"XML Generation"},{"line_number":153,"context_line":"--------------"},{"line_number":154,"context_line":""},{"line_number":155,"context_line":"Indirect mode SR-IOV does not encode the pci address in the libvirt xml."},{"line_number":156,"context_line":"The XML update logic that was introduced in the multiple port bindings"},{"line_number":157,"context_line":"feature is sufficent to enable the indirect use case."},{"line_number":158,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_76ab0c90","line":155,"range":{"start_line":155,"start_character":41,"end_line":155,"end_character":44},"updated":"2018-12-10 18:19:39.000000000","message":"PCI","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":152,"context_line":"XML Generation"},{"line_number":153,"context_line":"--------------"},{"line_number":154,"context_line":""},{"line_number":155,"context_line":"Indirect mode SR-IOV does not encode the pci address in the libvirt xml."},{"line_number":156,"context_line":"The XML update logic that was introduced in the multiple port bindings"},{"line_number":157,"context_line":"feature is sufficent to enable the indirect use case."},{"line_number":158,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_96aee89d","line":155,"range":{"start_line":155,"start_character":68,"end_line":155,"end_character":71},"updated":"2018-12-10 18:19:39.000000000","message":"XML","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":156,"context_line":"The XML update logic that was introduced in the multiple port bindings"},{"line_number":157,"context_line":"feature is sufficent to enable the indirect use case."},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Direct mode SR-IOV does encode the pci address in the libvirt xml, however,"},{"line_number":160,"context_line":"as the SR-IOV devices will be detached before migration and attached after"},{"line_number":161,"context_line":"migration no xml updates will be required."},{"line_number":162,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_36c91479","line":159,"range":{"start_line":159,"start_character":35,"end_line":159,"end_character":38},"updated":"2018-12-10 18:19:39.000000000","message":"PCI","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":156,"context_line":"The XML update logic that was introduced in the multiple port bindings"},{"line_number":157,"context_line":"feature is sufficent to enable the indirect use case."},{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Direct mode SR-IOV does encode the pci address in the libvirt xml, however,"},{"line_number":160,"context_line":"as the SR-IOV devices will be detached before migration and attached after"},{"line_number":161,"context_line":"migration no xml updates will be required."},{"line_number":162,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f6be3ccd","line":159,"range":{"start_line":159,"start_character":62,"end_line":159,"end_character":65},"updated":"2018-12-10 18:19:39.000000000","message":"XML","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":158,"context_line":""},{"line_number":159,"context_line":"Direct mode SR-IOV does encode the pci address in the libvirt xml, however,"},{"line_number":160,"context_line":"as the SR-IOV devices will be detached before migration and attached after"},{"line_number":161,"context_line":"migration no xml updates will be required."},{"line_number":162,"context_line":""},{"line_number":163,"context_line":""},{"line_number":164,"context_line":"Alternatives"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_16bad8d8","line":161,"range":{"start_line":161,"start_character":13,"end_line":161,"end_character":16},"updated":"2018-12-10 18:19:39.000000000","message":"XML","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":171,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":172,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":173,"context_line":""},{"line_number":174,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":175,"context_line":"  port and explicitly fail to migrate in this case with a new error."},{"line_number":176,"context_line":""},{"line_number":177,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_76dc6c15","line":174,"range":{"start_line":174,"start_character":21,"end_line":174,"end_character":24},"updated":"2018-12-10 18:19:39.000000000","message":"API","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":171,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":172,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":173,"context_line":""},{"line_number":174,"context_line":"* we could add a new api check to determine if an instance has an SR-IOV"},{"line_number":175,"context_line":"  port and explicitly fail to migrate in this case with a new error."},{"line_number":176,"context_line":""},{"line_number":177,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d6d9c026","line":174,"range":{"start_line":174,"start_character":2,"end_line":174,"end_character":4},"updated":"2018-12-10 18:19:39.000000000","message":"We","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":179,"context_line":"-----------------"},{"line_number":180,"context_line":""},{"line_number":181,"context_line":"It is expected that no data model changes should be required as the existing"},{"line_number":182,"context_line":"vif object in the migration_data object should be able to store the associated"},{"line_number":183,"context_line":"PCI address info. if this is not the case a small extension to those objects"},{"line_number":184,"context_line":"will be required for this info."},{"line_number":185,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_76f54c85","line":182,"range":{"start_line":182,"start_character":0,"end_line":182,"end_character":3},"updated":"2018-12-10 18:19:39.000000000","message":"VIF? Or ``vif``?","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":179,"context_line":"-----------------"},{"line_number":180,"context_line":""},{"line_number":181,"context_line":"It is expected that no data model changes should be required as the existing"},{"line_number":182,"context_line":"vif object in the migration_data object should be able to store the associated"},{"line_number":183,"context_line":"PCI address info. if this is not the case a small extension to those objects"},{"line_number":184,"context_line":"will be required for this info."},{"line_number":185,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_96f8287d","line":182,"range":{"start_line":182,"start_character":18,"end_line":182,"end_character":32},"updated":"2018-12-10 18:19:39.000000000","message":"``migration_data``","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":183,"context_line":"PCI address info. if this is not the case a small extension to those objects"},{"line_number":184,"context_line":"will be required for this info."},{"line_number":185,"context_line":""},{"line_number":186,"context_line":"A flavour extra spec and/or an image property may be desirable to opt into"},{"line_number":187,"context_line":"the direct mode auto hotplug behaviour"},{"line_number":188,"context_line":""},{"line_number":189,"context_line":""},{"line_number":190,"context_line":"REST API impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_9606e872","line":187,"range":{"start_line":186,"start_character":0,"end_line":187,"end_character":38},"updated":"2018-12-10 18:19:39.000000000","message":"If we do this how would we handle the absence of such a extra spec/image property - refuse to live migrate? Do we need this level of complexity? How about just \"forcing\" this on all direct mode live migration (with the caveat of documenting the crap out of the loss of network behaviour, see below)?","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"266dee2324e3067432368c8d9cd2bb862fc4cdc6","unresolved":false,"context_lines":[{"line_number":183,"context_line":"PCI address info. if this is not the case a small extension to those objects"},{"line_number":184,"context_line":"will be required for this info."},{"line_number":185,"context_line":""},{"line_number":186,"context_line":"A flavour extra spec and/or an image property may be desirable to opt into"},{"line_number":187,"context_line":"the direct mode auto hotplug behaviour"},{"line_number":188,"context_line":""},{"line_number":189,"context_line":""},{"line_number":190,"context_line":"REST API impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_c8cfd152","line":187,"range":{"start_line":186,"start_character":0,"end_line":187,"end_character":38},"in_reply_to":"3f79a3b5_9606e872","updated":"2018-12-11 10:28:23.000000000","message":"++ Joining on Artom\u0027s comment here.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":183,"context_line":"PCI address info. if this is not the case a small extension to those objects"},{"line_number":184,"context_line":"will be required for this info."},{"line_number":185,"context_line":""},{"line_number":186,"context_line":"A flavour extra spec and/or an image property may be desirable to opt into"},{"line_number":187,"context_line":"the direct mode auto hotplug behaviour"},{"line_number":188,"context_line":""},{"line_number":189,"context_line":""},{"line_number":190,"context_line":"REST API impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b507af20","line":187,"range":{"start_line":186,"start_character":0,"end_line":187,"end_character":38},"in_reply_to":"3f79a3b5_c8cfd152","updated":"2018-12-20 17:42:22.000000000","message":"ok ill remove this.\n\ni originally had planned to abort the migration if not set but i also had considered forcing the the detach behavior.\n\nthe flavor extraspec/image metadata was intended to allow operators/tenats to choose.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":205,"context_line":"Other end user impact"},{"line_number":206,"context_line":"---------------------"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"Users of direct mode SR-IOV should be aware that auto hotplugging"},{"line_number":209,"context_line":"is not transparent to the guest in exactly the same way that"},{"line_number":210,"context_line":"suspend is not transparent today."},{"line_number":211,"context_line":""},{"line_number":212,"context_line":""},{"line_number":213,"context_line":"Performance Impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b624441b","line":210,"range":{"start_line":208,"start_character":0,"end_line":210,"end_character":33},"updated":"2018-12-10 18:19:39.000000000","message":"This needs to be made super explicit in release notes, admin guides, and any other appropriate documentation, because as far as I know this is the first time we\u0027re breaking the \"live migration is transparent to the workload (unless it breaks)\" precedent/paradigm.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":205,"context_line":"Other end user impact"},{"line_number":206,"context_line":"---------------------"},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"Users of direct mode SR-IOV should be aware that auto hotplugging"},{"line_number":209,"context_line":"is not transparent to the guest in exactly the same way that"},{"line_number":210,"context_line":"suspend is not transparent today."},{"line_number":211,"context_line":""},{"line_number":212,"context_line":""},{"line_number":213,"context_line":"Performance Impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_104a3159","line":210,"range":{"start_line":208,"start_character":0,"end_line":210,"end_character":33},"in_reply_to":"3f79a3b5_b624441b","updated":"2018-12-20 17:42:22.000000000","message":"well as far as im aware hyperv already does this internally.\nthere is also precedent for this in suspend/snapshot.\n\nwe don\u0027t document that suspend is not transparent in the libvirt driver where we also detach the pci devices on suspend and attach them on restore. in all other virt drivers that im aware of suspend is transparent to the guest bar the change in the clock.\n\nbut yes we should document it as its as suprising as the suspend behavior being non transparent.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":214,"context_line":"------------------"},{"line_number":215,"context_line":""},{"line_number":216,"context_line":"This should not significantly impact the performance of a live migration."},{"line_number":217,"context_line":"A minor overhead will be incurred in claiming the resource and updating the xml"},{"line_number":218,"context_line":""},{"line_number":219,"context_line":"Other deployer impact"},{"line_number":220,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_762a2ce5","line":217,"range":{"start_line":217,"start_character":76,"end_line":217,"end_character":79},"updated":"2018-12-10 18:19:39.000000000","message":"XML","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":219,"context_line":"Other deployer impact"},{"line_number":220,"context_line":"---------------------"},{"line_number":221,"context_line":""},{"line_number":222,"context_line":"If a new flavour extra spec or image property is added to enable this"},{"line_number":223,"context_line":"functionality then a deployer will have to update there flavours/images"},{"line_number":224,"context_line":"to use this feature."},{"line_number":225,"context_line":""},{"line_number":226,"context_line":"Similarly, if a config option is added instead they will have to modify"},{"line_number":227,"context_line":"their config."},{"line_number":228,"context_line":""},{"line_number":229,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"},{"line_number":230,"context_line":"defaults for direct mode SR-IOV. SR-IOV live migration will be enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_967b48f0","line":227,"range":{"start_line":222,"start_character":0,"end_line":227,"end_character":13},"updated":"2018-12-10 18:19:39.000000000","message":"Depending on the outcome of the discussion on L187, drop these.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":226,"context_line":"Similarly, if a config option is added instead they will have to modify"},{"line_number":227,"context_line":"their config."},{"line_number":228,"context_line":""},{"line_number":229,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"},{"line_number":230,"context_line":"defaults for direct mode SR-IOV. SR-IOV live migration will be enabled"},{"line_number":231,"context_line":"if both the source and dest node support it."},{"line_number":232,"context_line":""},{"line_number":233,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_d6ba206e","line":230,"range":{"start_line":229,"start_character":63,"end_line":230,"end_character":31},"updated":"2018-12-10 18:19:39.000000000","message":"And do what? Refuse to live migrate SR-IOV direct mode unless that config option/extra spec/image property is present? I\u0027m kinda coming back to the idea of \"forcing\" it on everyone.\n\nIf we do go the \"make it configurable\" route, we need to decide in the spec how that\u0027ll be done, and include that in the \"Proposed change\" section. I\u0027m also thinking of the N/N+1 live migration case, see below.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"eb2617edc18ea364b8bc57469934c7b0de4f7411","unresolved":false,"context_lines":[{"line_number":226,"context_line":"Similarly, if a config option is added instead they will have to modify"},{"line_number":227,"context_line":"their config."},{"line_number":228,"context_line":""},{"line_number":229,"context_line":"As auto hotplugging is not transparent to the guest it will be disabled by"},{"line_number":230,"context_line":"defaults for direct mode SR-IOV. SR-IOV live migration will be enabled"},{"line_number":231,"context_line":"if both the source and dest node support it."},{"line_number":232,"context_line":""},{"line_number":233,"context_line":"Developer impact"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_b43cbcbd","line":230,"range":{"start_line":229,"start_character":63,"end_line":230,"end_character":31},"in_reply_to":"3f79a3b5_d6ba206e","updated":"2018-12-17 12:45:52.000000000","message":"i was originally thinking of a key set in the neutron  port binding profile but i recently discovered that setting values in the binding profile is admin only by default so it defeats the idea of allowing the teant to decide.\n\nAn image metadata value would work\n\ni dont really want this to be a config thing as i do not think we shoud alter this type of behavior via config.\n\nbut ok i can drop this and simply state we will enbable if both nodes support it.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_566b70ec","line":238,"updated":"2018-12-10 18:19:39.000000000","message":"On the NUMA live migration spec, Sylvain made a good point that refusing live migrations in the N/N+1 case is a bad idea, as operatora depend on it to move workloads off compute nodes to be upgraded. So I think we need a truth table here, similar to what I did on the NUMA live migration spec. This is also why I don\u0027t really want to add the complexity of the config option for the auto hotplugging, as it\u0027ll turn the truth table from 2 x 2 to 2 x 2 x 2 (I may be overthinking it and we don\u0027t really need it, but having all the possibilities spelled out would help reassure me).","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"dfd5e7cf_c5c73e45","line":238,"in_reply_to":"3f79a3b5_30f64dd8","updated":"2019-01-10 16:36:12.000000000","message":"\u003e the current behavior is that without this change if using macvtap\n \u003e we move the instance to the destination host and raises an\n \u003e exception in post migrate before we update the host in the nova db.\n \u003e this result in nova thinking the vm is still running on the source\n \u003e host but its actually running on the destination.\n\nWhat raises the exception? This is just ignorance on my part, as I\u0027ve neither played with SRIOV live migration of looked at the code in detail, but in your problem description for indirect mode you only mention resource claims. How would failing to claim the resource raise an exception?","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"e6428b30a1cf240bc415b901b30834d0d78db160","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_af5f89dc","line":238,"in_reply_to":"3f79a3b5_342f4cf5","updated":"2018-12-17 13:20:23.000000000","message":"\u003e we could have the conductor check fail the migration if only\n \u003e one of the two selected host support the feature.\n\nThat\u0027s what I had originially for NUMA live migration, but it got changed because blocking it outright isn\u0027t cool for operators who rely on NUMA live migration (in its current half-broken state) to move workloads off compute nodes about to be updated, hence we need to support the N/N+1 scenario.\n\nI\u0027m wondering whether something similar is needed here - and that\u0027s why I think a problem description split into Intro, Direct, and Indirect (see my comment on L19) would be useful. Obviously if both direct and indirect are 100% broken with exceptions all over the place, then blocking unless both nodes support the new flow makes sense. If there\u0027s a specific case where a SRIOV live migration in its current state can go through successfully, and some operators (for whatever black magic reasons) rely on that, I don\u0027t think blocking is correct.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"266dee2324e3067432368c8d9cd2bb862fc4cdc6","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_e8708dda","line":238,"in_reply_to":"3f79a3b5_566b70ec","updated":"2018-12-11 10:28:23.000000000","message":"IMO refusing to do something which is wrong 90% of the time is the correct way to go.\noperators should not rely on this buggy behaviour. (dare i say it is a bug that it was not blocked to begin with?)\n\nOn another note,\nI tried to perform a macVtap live migration with the current upstream code and it seems the migration fails during VIF plug on destination\nas neutronV2 \u0027bind_ports_to_host()\u0027 does not preserve the \u0027vnic_type\u0027(direct) and \u0027profile\u0027. it performs a fallback to a normal \u0027vnic_type\u0027.\nSo i am not entirely sure we need to bother ourselves with preserving old behaviour. \nEven if that worked (os_vif.plug()) im not sure the performance degradation is something the operator planned when the migration was triggered.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_30f64dd8","line":238,"in_reply_to":"3f79a3b5_8addbfdb","updated":"2018-12-20 17:42:22.000000000","message":"with the numa aware migration spec when not using cpu pinning the migration not only worked a very hi percentage of the time but was also correct when it scuceeded. if you use pinning yes it was not respecting the pinning policy but it still was posible to do the migration.\n\nthe difference in this case is sriov migration with the libvirt driver has never actully worked. as such blocking it is not a regression. infact it prevents a bug that happens if you attempt to migrate macvtap instance today.\n\nthe current behavior is that without this change if using macvtap we move the instance to the destination host and raises an exception in post migrate before we update the host in the nova db. this result in nova thinking the vm is still running on the source host but its actually running on the destination.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"e0a2f9e62c8122549ef7be679ce00ee956bf8ab5","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_8addbfdb","line":238,"in_reply_to":"3f79a3b5_af5f89dc","updated":"2018-12-18 14:02:29.000000000","message":"Digging in the code abit, the bug in neutronv2.api.bind_ports_to_host()[1] is there from day one, about a year ago [2].\nNow if live-migration did not use the multiple port binding feature, the binding of the port on the destination is done in neutronv2.api.setup_networks_on_host()[3], the binding will only work if the same PCI function is available on the destination host (hopefully connected to the same physical network).\n\n\nSupport for PCI resource claiming and port binding profile update is lacking at nova compute level(not only in some virt drivers), live migration will fail in most cases and will lead to incorrect resource tracking when it succeeds under very specific conditions.\n\n\nI have no major objection on keeping old behavior if all conditions are not met, however i feel like it will bring little benefit.\n\n\n[1]https://github.com/openstack/nova/blob/5d748001c2ffb1197ff991ca69a95865087f1010/nova/network/neutronv2/api.py#L1232\n[2] https://github.com/openstack/nova/commit/750e065d6d137715dd03d5c90623145177f64cbb\n[3] https://github.com/openstack/nova/blob/5d748001c2ffb1197ff991ca69a95865087f1010/nova/network/neutronv2/api.py#L369","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"eb2617edc18ea364b8bc57469934c7b0de4f7411","unresolved":false,"context_lines":[{"line_number":235,"context_line":""},{"line_number":236,"context_line":"None"},{"line_number":237,"context_line":""},{"line_number":238,"context_line":"Upgrade impact"},{"line_number":239,"context_line":"--------------"},{"line_number":240,"context_line":""},{"line_number":241,"context_line":"This feature may aid upgrade of hosts with SR-IOV enabled"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_342f4cf5","line":238,"in_reply_to":"3f79a3b5_e8708dda","updated":"2018-12-17 12:45:52.000000000","message":"@artom i dont think we need a truth table. its simple\nif both nodes support the feture they get new behavior\nif either of them does not the get the old behavior.\nthis is intended to be identical to how we did cross version support for using multiple port bindings which is what i argured the numa migration spec should have stated also.\nhttps://specs.openstack.org/openstack/nova-specs/specs/rocky/implemented/neutron-new-port-binding-api.html#upgrade-impact\n\n@adrian it would be incorrect to encoded the behavior of one driver in the api. my understanding is that the hyperv driver already support pci live migration using the detach attach logic i have described. in the hyperv case this is handled by hyperv itself and not nova. As such we cannot and should not refuse to live migrate instance with sriov port at the api. we could add a driver check in the libvirt driver but that would only work if at least on of the node supported it.\n\nThe fact that the vnic_type and profile were not preseved was/is a bug not the intended behavior so that is not the behavior i was suggesting falling back too. \n\n\nwe could have the conductor check fail the migration if only\none of the two selected host support the feature. i do not believe we retry live-migrations however so this would try one host pair and then fail/pass unless we added retry logic which i think is not warrented. \n\nwe could keep retrying if the source node supported the feature to find a dest node that also did but this is technical debt that would only be usefaul for one release and i want to avoid adding it by only supporting it if you have 2 stien nodes.\n\ni will add a statement to the effect that if only 1 of the 2 nodes support this feature we will abort the migration to keep the conductor check simple. and ensure we do not have to implement support for Rocket\u003c-\u003eStein live migration and only support Stein+","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":276,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":277,"context_line":""},{"line_number":278,"context_line":"This spec has no dependencies but intends to collaborate with the"},{"line_number":279,"context_line":"implementation of NUMA aware live migration [1_]"},{"line_number":280,"context_line":""},{"line_number":281,"context_line":"Note that modification to the sriovnicswitch ml2 driver may"},{"line_number":282,"context_line":"be required to support multiple port binings. This work if needed"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_963ee8df","line":279,"range":{"start_line":279,"start_character":44,"end_line":279,"end_character":48},"updated":"2018-12-10 18:19:39.000000000","message":"[1]_","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"11c82f529263219457faf120c5a98dbce1bf0982","unresolved":false,"context_lines":[{"line_number":279,"context_line":"implementation of NUMA aware live migration [1_]"},{"line_number":280,"context_line":""},{"line_number":281,"context_line":"Note that modification to the sriovnicswitch ml2 driver may"},{"line_number":282,"context_line":"be required to support multiple port binings. This work if needed"},{"line_number":283,"context_line":"is out of scope of this spec and will be tracked using Neutron"},{"line_number":284,"context_line":"RFE bugs and/or specs as required."},{"line_number":285,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_70c1d482","line":282,"range":{"start_line":282,"start_character":37,"end_line":282,"end_character":44},"updated":"2018-12-10 16:00:48.000000000","message":"bindings","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":299,"context_line":"Documentation Impact"},{"line_number":300,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":301,"context_line":""},{"line_number":302,"context_line":"operator docs will need to be updated to describe how to enable the new"},{"line_number":303,"context_line":"feature. End User docs should be updated to document that direct mode"},{"line_number":304,"context_line":"auto-attach is not transparent."},{"line_number":305,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_5654d022","line":302,"range":{"start_line":302,"start_character":0,"end_line":302,"end_character":8},"updated":"2018-12-10 18:19:39.000000000","message":"Operator","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":300,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":301,"context_line":""},{"line_number":302,"context_line":"operator docs will need to be updated to describe how to enable the new"},{"line_number":303,"context_line":"feature. End User docs should be updated to document that direct mode"},{"line_number":304,"context_line":"auto-attach is not transparent."},{"line_number":305,"context_line":""},{"line_number":306,"context_line":"References"},{"line_number":307,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_161f386e","line":304,"range":{"start_line":303,"start_character":9,"end_line":304,"end_character":31},"updated":"2018-12-10 18:19:39.000000000","message":"End users can\u0027t do live migrations with our default policy, right? So I don\u0027t think we need this, but as I\u0027ve mentioned above, the auto hotplugging not being transparent needs to be shouted from rooftops.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"23551d1670bbeed219578989857286833ae82e96","unresolved":false,"context_lines":[{"line_number":300,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":301,"context_line":""},{"line_number":302,"context_line":"operator docs will need to be updated to describe how to enable the new"},{"line_number":303,"context_line":"feature. End User docs should be updated to document that direct mode"},{"line_number":304,"context_line":"auto-attach is not transparent."},{"line_number":305,"context_line":""},{"line_number":306,"context_line":"References"},{"line_number":307,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_70c7e567","line":304,"range":{"start_line":303,"start_character":9,"end_line":304,"end_character":31},"in_reply_to":"3f79a3b5_161f386e","updated":"2018-12-20 17:42:22.000000000","message":"sure but as i also pointed out we dont do that for suspend which is intended to be entirely transparent and is in all drivers except libvirt.\n\nDocumenting it in the end user docs was for the image metadata property so they knew to set it if it was required.\n\nThis is not need anymore.","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"66511c11adcb8d7ae1e97ea3d2843065504a6b7c","unresolved":false,"context_lines":[{"line_number":306,"context_line":"References"},{"line_number":307,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":308,"context_line":""},{"line_number":309,"context_line":".. _[0]: https://specs.openstack.org/openstack/nova-specs/specs/juno/implemented/pci-passthrough-sriov.html"},{"line_number":310,"context_line":".. _[1]: https://review.openstack.org/#/c/599587/2/specs/stein/approved/numa-aware-live-migration.rst"},{"line_number":311,"context_line":".. _[2]: https://github.com/openstack/nova/blob/2f635fa914884c91f3b8cc78cda5154dd3b43305/nova/virt/libvirt/driver.py#L2912-L2920"},{"line_number":312,"context_line":".. _[3]: https://github.com/openstack/nova/blob/2f635fa914884c91f3b8cc78cda5154dd3b43305/nova/virt/libvirt/driver.py#L2929-L2932"},{"line_number":313,"context_line":""},{"line_number":314,"context_line":"History"},{"line_number":315,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":7,"id":"3f79a3b5_f4f662d6","line":312,"range":{"start_line":309,"start_character":0,"end_line":312,"end_character":128},"updated":"2018-12-10 18:19:39.000000000","message":"nit: if you remove the colons the links will show up properly","commit_id":"4e15a108b69139786edd9a62af0ec7739ea0bf18"},{"author":{"_account_id":8768,"name":"Chris Friesen","email":"chris.friesen@windriver.com","username":"cbf123"},"change_message_id":"255dcc78e623b78dc6f9b479f083c2b4b82d8fde","unresolved":false,"context_lines":[{"line_number":96,"context_line":"As the provider of a cloud with a hardware offloaded vSwitch that leverages"},{"line_number":97,"context_line":"indirect mode SR-IOV, I want to offer the performance it enables to my"},{"line_number":98,"context_line":"customers but also desire the flexibility to be able to transparently migrate"},{"line_number":99,"context_line":"guests without disrupting traffic to enable maintenance."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Proposed change"},{"line_number":102,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3f79a3b5_377db9c2","line":99,"range":{"start_line":99,"start_character":7,"end_line":99,"end_character":33},"updated":"2018-12-20 20:15:23.000000000","message":"there\u0027s still going to be some disruption during live migration, maybe this should talk about minimal disruption rather than no disruption?","commit_id":"20d5e1e8158cab5e0300812b29d4bbc185913e0c"},{"author":{"_account_id":8768,"name":"Chris Friesen","email":"chris.friesen@windriver.com","username":"cbf123"},"change_message_id":"255dcc78e623b78dc6f9b479f083c2b4b82d8fde","unresolved":false,"context_lines":[{"line_number":118,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":119,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":120,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":121,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":122,"context_line":"success and failure case we do not need to rely on the"},{"line_number":123,"context_line":"``update_available_resource`` periodic task to heal the allocations/claims."},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"SR-IOV Mode"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3f79a3b5_37a61957","line":123,"range":{"start_line":121,"start_character":22,"end_line":123,"end_character":75},"updated":"2018-12-20 20:15:23.000000000","message":"Agreed, this is the way to go.","commit_id":"20d5e1e8158cab5e0300812b29d4bbc185913e0c"},{"author":{"_account_id":28714,"name":"Adrian Chiris","email":"adrianc@nvidia.com","username":"adrianc"},"change_message_id":"366b09872fb11a751ddd605b6dab98fe19c90b20","unresolved":false,"context_lines":[{"line_number":118,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":119,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":120,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":121,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":122,"context_line":"success and failure case we do not need to rely on the"},{"line_number":123,"context_line":"``update_available_resource`` periodic task to heal the allocations/claims."},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"SR-IOV Mode"}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffd0ebdf_12ae412e","line":123,"range":{"start_line":121,"start_character":22,"end_line":123,"end_character":75},"in_reply_to":"1f769fc5_50ff9233","updated":"2019-01-03 14:32:51.000000000","message":"Indeed, the periodic task will still heal what it heals today, this includes PCI resources.","commit_id":"20d5e1e8158cab5e0300812b29d4bbc185913e0c"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"fda4d98f5037220a4035422dcf5ca2b3cc605098","unresolved":false,"context_lines":[{"line_number":118,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":119,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":120,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":121,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":122,"context_line":"success and failure case we do not need to rely on the"},{"line_number":123,"context_line":"``update_available_resource`` periodic task to heal the allocations/claims."},{"line_number":124,"context_line":""},{"line_number":125,"context_line":""},{"line_number":126,"context_line":"SR-IOV Mode"}],"source_content_type":"text/x-rst","patch_set":9,"id":"1f769fc5_50ff9233","line":123,"range":{"start_line":121,"start_character":22,"end_line":123,"end_character":75},"in_reply_to":"3f79a3b5_37a61957","updated":"2019-01-02 16:14:59.000000000","message":"I also agree that being proactive is the way to go for these things. Explicit, predictable.\n\nHowever, we\u0027ve had it made clear time and again that being able to heal from unexpected things (net partition from placement?) in the periodic is handy.","commit_id":"20d5e1e8158cab5e0300812b29d4bbc185913e0c"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":7,"context_line":"Neutron SR-IOV Port Live Migration"},{"line_number":8,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV by [0]_, live migration was not"},{"line_number":13,"context_line":"directly addressed as part of the proposed changes. As a result while"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_6a7660a7","line":10,"range":{"start_line":10,"start_character":0,"end_line":10,"end_character":79},"updated":"2019-01-10 14:55:52.000000000","message":"valid","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":9,"context_line":""},{"line_number":10,"context_line":"https://blueprints.launchpad.net/nova/+spec/libvirt-neutron-sriov-livemigration"},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"When nova was extended to support SR-IOV by [0]_, live migration was not"},{"line_number":13,"context_line":"directly addressed as part of the proposed changes. As a result while"},{"line_number":14,"context_line":"live migration with SR-IOV is technically feasible, it remains unsupported"},{"line_number":15,"context_line":"by the libvirt driver. This spec seeks to address this gap in live migration"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_aab8c829","line":12,"range":{"start_line":12,"start_character":41,"end_line":12,"end_character":43},"updated":"2019-01-10 14:55:52.000000000","message":"nit: drop this ([0]_ is a cross-reference. You just leave it beside whatever you were talking about)","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":27,"context_line":"*     resource claims"},{"line_number":28,"context_line":""},{"line_number":29,"context_line":"NUMA affinity is out of the scope of this spec and will be addressed"},{"line_number":30,"context_line":"separately by [1]_."},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"The SR-IOV mode of a neutron port directly impacts how a live migration"},{"line_number":33,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_8a870462","line":30,"range":{"start_line":30,"start_character":11,"end_line":30,"end_character":13},"updated":"2019-01-10 14:55:52.000000000","message":"ditto","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":32,"context_line":"The SR-IOV mode of a neutron port directly impacts how a live migration"},{"line_number":33,"context_line":"can be done. This spec will focus on the two categories of SR-IOV primarily,"},{"line_number":34,"context_line":"direct passthrough (``vnic_type\u003ddirect|direct-physical``) and indirect"},{"line_number":35,"context_line":"passthrough (``vnic_type\u003dmacvtap|virtio-forwarder``). For simplicity, direct"},{"line_number":36,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":37,"context_line":"of the spec."},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":40,"context_line":"performance is achieved at the cost of exposing the guest to the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_ca718c0e","line":37,"range":{"start_line":35,"start_character":53,"end_line":37,"end_character":12},"updated":"2019-01-10 15:53:46.000000000","message":"thank you.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":37,"context_line":"of the spec."},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":40,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":41,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":42,"context_line":"state, direct mode SR-IOV, cannot be conventionally live migrated."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_aa91a89b","line":39,"updated":"2019-01-10 14:55:52.000000000","message":"nit: SR-IOV, (comma)","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":36,"context_line":"mode and indirect mode are used instead of ``vnic_type`` in the rest"},{"line_number":37,"context_line":"of the spec."},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":40,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":41,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":42,"context_line":"state, direct mode SR-IOV, cannot be conventionally live migrated."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_65e532a8","line":39,"in_reply_to":"dfd5e7cf_aa91a89b","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":39,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":40,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":41,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":42,"context_line":"state, direct mode SR-IOV, cannot be conventionally live migrated."},{"line_number":43,"context_line":"This spec will provide a workaround to enable this configuration."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_0a51d462","line":42,"range":{"start_line":42,"start_character":25,"end_line":42,"end_character":27},"updated":"2019-01-10 15:53:46.000000000","message":"remove this comma","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":39,"context_line":"When a device is exposed to a guest via direct mode SR-IOV maximum"},{"line_number":40,"context_line":"performance is achieved at the cost of exposing the guest to the"},{"line_number":41,"context_line":"hardware state. Since there is no standard mechanism to copy the hardware"},{"line_number":42,"context_line":"state, direct mode SR-IOV, cannot be conventionally live migrated."},{"line_number":43,"context_line":"This spec will provide a workaround to enable this configuration."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_c5d65ec4","line":42,"range":{"start_line":42,"start_character":25,"end_line":42,"end_character":27},"in_reply_to":"dfd5e7cf_0a51d462","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":44,"context_line":""},{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":46,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_6ac74014","line":47,"range":{"start_line":47,"start_character":39,"end_line":47,"end_character":46},"updated":"2019-01-10 15:53:46.000000000","message":"explicit","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":44,"context_line":""},{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":46,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_05d546b9","line":47,"range":{"start_line":47,"start_character":39,"end_line":47,"end_character":46},"in_reply_to":"dfd5e7cf_6ac74014","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":46,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":51,"context_line":"          offload state or Tx/Rx queues that are implemented in hardware"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_ca8c2c81","line":48,"range":{"start_line":48,"start_character":53,"end_line":48,"end_character":58},"updated":"2019-01-10 14:55:52.000000000","message":"nit: SR-IOV","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":45,"context_line":"Hardware state transfer is a property of SR-IOV live migration that cannot"},{"line_number":46,"context_line":"be addressed by OpenStack, as such this spec does not intend to copy hardware"},{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":51,"context_line":"          offload state or Tx/Rx queues that are implemented in hardware"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e5d122cf","line":48,"range":{"start_line":48,"start_character":53,"end_line":48,"end_character":58},"in_reply_to":"dfd5e7cf_ca8c2c81","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":51,"context_line":"          offload state or Tx/Rx queues that are implemented in hardware"},{"line_number":52,"context_line":"          which is not software programmable via the hypervisor e.g. mac"},{"line_number":53,"context_line":"          address is not considered hardware state in this context as"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e5ccc2bd","line":50,"range":{"start_line":50,"start_character":55,"end_line":50,"end_character":58},"updated":"2019-01-10 16:36:12.000000000","message":"NIC","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":47,"context_line":"state. Copying hardware state requires explcit support at the hardware,"},{"line_number":48,"context_line":"driver and hypervisor level which does not exist for sriov devices."},{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":51,"context_line":"          offload state or Tx/Rx queues that are implemented in hardware"},{"line_number":52,"context_line":"          which is not software programmable via the hypervisor e.g. mac"},{"line_number":53,"context_line":"          address is not considered hardware state in this context as"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_90d9d423","line":50,"range":{"start_line":50,"start_character":55,"end_line":50,"end_character":58},"in_reply_to":"dfd5e7cf_e5ccc2bd","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":49,"context_line":""},{"line_number":50,"context_line":".. note:: hardware state in this context refers to any nic state such as"},{"line_number":51,"context_line":"          offload state or Tx/Rx queues that are implemented in hardware"},{"line_number":52,"context_line":"          which is not software programmable via the hypervisor e.g. mac"},{"line_number":53,"context_line":"          address is not considered hardware state in this context as"},{"line_number":54,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":55,"context_line":"          host level interface."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_eada50e7","line":52,"range":{"start_line":52,"start_character":69,"end_line":52,"end_character":72},"updated":"2019-01-10 15:53:46.000000000","message":"here and throughout..  s/mac address/MAC address/","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"94f0a6f55cf0f97c53b296ccc294843fe171a2ba","unresolved":false,"context_lines":[{"line_number":54,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":55,"context_line":"          host level interface."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"},{"line_number":58,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":59,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_487a8388","line":57,"range":{"start_line":57,"start_character":12,"end_line":57,"end_character":25},"updated":"2019-01-10 17:05:49.000000000","message":"nit: SR-IOV indirect mode","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":54,"context_line":"          libvirt/qemu can set the mac of an SR-IOV device via a standard"},{"line_number":55,"context_line":"          host level interface."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"},{"line_number":58,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":59,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_b0d658f1","line":57,"range":{"start_line":57,"start_character":12,"end_line":57,"end_character":25},"in_reply_to":"dfd5e7cf_487a8388","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":57,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"},{"line_number":58,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":59,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"},{"line_number":61,"context_line":"therefore allows migration of guests without any workarounds."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_658f7323","line":60,"range":{"start_line":60,"start_character":0,"end_line":60,"end_character":11},"updated":"2019-01-10 15:53:46.000000000","message":"I think saying \"From a guest perspective, the SR-IOV interfaces are exposed as virtual NICs\" is easier to read and understand. Agree?","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":57,"context_line":"For indirect mode SR-IOV, the SR-IOV device is exposed via a software"},{"line_number":58,"context_line":"mediation layer such as macvtap + kernel vhost, vhost-user or vhost-vfio."},{"line_number":59,"context_line":"From a guest perspective, the SR-IOV interfaces are exposed as a standard"},{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"},{"line_number":61,"context_line":"therefore allows migration of guests without any workarounds."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_25f16a6c","line":60,"range":{"start_line":60,"start_character":0,"end_line":60,"end_character":11},"in_reply_to":"dfd5e7cf_658f7323","updated":"2019-01-10 18:25:00.000000000","message":"yes.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":67,"context_line":"device an exception is raised after the VM has been sucessfully migrated to the"},{"line_number":68,"context_line":"destination in the post migration cleanup."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_aae68843","line":65,"range":{"start_line":65,"start_character":67,"end_line":65,"end_character":75},"updated":"2019-01-10 14:55:52.000000000","message":"nit: , however,","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":67,"context_line":"device an exception is raised after the VM has been sucessfully migrated to the"},{"line_number":68,"context_line":"destination in the post migration cleanup."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_45cb4e1d","line":65,"range":{"start_line":65,"start_character":67,"end_line":65,"end_character":75},"in_reply_to":"dfd5e7cf_aae68843","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"},{"line_number":61,"context_line":"therefore allows migration of guests without any workarounds."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":67,"context_line":"device an exception is raised after the VM has been sucessfully migrated to the"},{"line_number":68,"context_line":"destination in the post migration cleanup."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_45326ea9","line":68,"range":{"start_line":63,"start_character":0,"end_line":68,"end_character":42},"updated":"2019-01-10 16:36:12.000000000","message":"OK, so we do need resource claims for indirect mode as well.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":60,"context_line":"virtio-net-PCI device and no hardware state is observed. Indirect mode SR-IOV"},{"line_number":61,"context_line":"therefore allows migration of guests without any workarounds."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The main gap in SR-IOV live migration support today is resource claims."},{"line_number":64,"context_line":"As mentioned in the introduction it is technically possible to live migrate"},{"line_number":65,"context_line":"a guest with an indirect mode SR-IOV device between two hosts today however"},{"line_number":66,"context_line":"when you do, resources are not correctly claimed. By not claiming the SR-IOV"},{"line_number":67,"context_line":"device an exception is raised after the VM has been sucessfully migrated to the"},{"line_number":68,"context_line":"destination in the post migration cleanup."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_5005ac79","line":68,"range":{"start_line":63,"start_character":0,"end_line":68,"end_character":42},"in_reply_to":"dfd5e7cf_45326ea9","updated":"2019-01-10 18:25:00.000000000","message":"yes","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":69,"context_line":""},{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":72,"context_line":"a free PCI device on the destination node, with the same PCI address as the"},{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_05c78747","line":72,"range":{"start_line":72,"start_character":41,"end_line":72,"end_character":42},"updated":"2019-01-10 15:53:46.000000000","message":"remove this comma","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":69,"context_line":""},{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":72,"context_line":"a free PCI device on the destination node, with the same PCI address as the"},{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e53b62e8","line":72,"range":{"start_line":72,"start_character":41,"end_line":72,"end_character":42},"in_reply_to":"dfd5e7cf_05c78747","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":72,"context_line":"a free PCI device on the destination node, with the same PCI address as the"},{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of two issues. First nova does not correctly claim the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_65dd3313","line":73,"range":{"start_line":73,"start_character":11,"end_line":73,"end_character":12},"updated":"2019-01-10 15:53:46.000000000","message":"remove this comma","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":70,"context_line":"When live migrating today, migration will also fail if the PCI mapping is"},{"line_number":71,"context_line":"required to change. Said another way, migration will only succeed if there is"},{"line_number":72,"context_line":"a free PCI device on the destination node, with the same PCI address as the"},{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of two issues. First nova does not correctly claim the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_45358ef6","line":73,"range":{"start_line":73,"start_character":11,"end_line":73,"end_character":12},"in_reply_to":"dfd5e7cf_65dd3313","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of two issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_cabaec08","line":76,"range":{"start_line":76,"start_character":31,"end_line":76,"end_character":36},"updated":"2019-01-10 14:55:52.000000000","message":"Firstly,","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":73,"context_line":"source node, that is connected to the same physnet and is on the same NUMA"},{"line_number":74,"context_line":"node."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"This is because of two issues. First nova does not correctly claim the"},{"line_number":77,"context_line":"SR-IOV device on the destination node and second, nova does not modify"},{"line_number":78,"context_line":"the guest XML to reflect the host PCI address on the destination."},{"line_number":79,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e59fe200","line":76,"range":{"start_line":76,"start_character":31,"end_line":76,"end_character":36},"in_reply_to":"dfd5e7cf_cabaec08","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":89,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":90,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":91,"context_line":"flexibility of live migration for maintenance. To that end, as an operator,"},{"line_number":92,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":93,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":94,"context_line":"understanding performance SLAs will not be met during the migration."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"As the provider of a cloud with a hardware offloaded vSwitch that leverages"},{"line_number":97,"context_line":"indirect mode SR-IOV, I want to offer the performance it enables to my"},{"line_number":98,"context_line":"customers but also desire the flexibility to be able to transparently migrate"},{"line_number":99,"context_line":"guests without disrupting traffic to enable maintenance."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Proposed change"},{"line_number":102,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_255c0b7d","line":99,"range":{"start_line":92,"start_character":56,"end_line":99,"end_character":56},"updated":"2019-01-10 15:53:46.000000000","message":"both of these use cases mention using indirect-mode SR-IOV interfaces, and yet under the proposed change section, it says indirect-mode requires no changes to existing code...","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":89,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":90,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":91,"context_line":"flexibility of live migration for maintenance. To that end, as an operator,"},{"line_number":92,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":93,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":94,"context_line":"understanding performance SLAs will not be met during the migration."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"As the provider of a cloud with a hardware offloaded vSwitch that leverages"},{"line_number":97,"context_line":"indirect mode SR-IOV, I want to offer the performance it enables to my"},{"line_number":98,"context_line":"customers but also desire the flexibility to be able to transparently migrate"},{"line_number":99,"context_line":"guests without disrupting traffic to enable maintenance."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Proposed change"},{"line_number":102,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_c893d3a2","line":99,"range":{"start_line":92,"start_character":56,"end_line":99,"end_character":56},"in_reply_to":"dfd5e7cf_255c0b7d","updated":"2019-01-10 18:25:00.000000000","message":"that comment was realy in relation to hardware state transfer.\n\nthe gap in resource claiming is common to indirect and direct mode and is the primery code cange for the feature.\n\ni could rephrase but i\u0027m not sure how to clarify really.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"b5e9642e0d9579f038b97c61abcf5a2e9a1c4907","unresolved":false,"context_lines":[{"line_number":89,"context_line":"that has a long peering time, I would like to be able to utilise"},{"line_number":90,"context_line":"direct mode SR-IOV to meet my performance SLAs but desire the"},{"line_number":91,"context_line":"flexibility of live migration for maintenance. To that end, as an operator,"},{"line_number":92,"context_line":"I am willing to use a bond in the guest to a vSwitch or indirect SR-IOV"},{"line_number":93,"context_line":"interface to facilitate migration and retain peering relationships while"},{"line_number":94,"context_line":"understanding performance SLAs will not be met during the migration."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"As the provider of a cloud with a hardware offloaded vSwitch that leverages"},{"line_number":97,"context_line":"indirect mode SR-IOV, I want to offer the performance it enables to my"},{"line_number":98,"context_line":"customers but also desire the flexibility to be able to transparently migrate"},{"line_number":99,"context_line":"guests without disrupting traffic to enable maintenance."},{"line_number":100,"context_line":""},{"line_number":101,"context_line":"Proposed change"},{"line_number":102,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_b31742ee","line":99,"range":{"start_line":92,"start_character":56,"end_line":99,"end_character":56},"in_reply_to":"dfd5e7cf_c893d3a2","updated":"2019-01-10 18:59:29.000000000","message":"Nah, it\u0027s cool.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":116,"context_line":"host PCI address. If the migration should fail after the destination resources"},{"line_number":117,"context_line":"have been claimed they must be released in the"},{"line_number":118,"context_line":"``rollback_live_migration_at_destination`` call. If the migration succeeds"},{"line_number":119,"context_line":"the source host sriov device will be freed in ``post_live_migration``"},{"line_number":120,"context_line":"(clean up source) and the state of claimed devices on the destination are"},{"line_number":121,"context_line":"updated to allocated. By proactively updating the resouce tracker in both the"},{"line_number":122,"context_line":"success and failure case we do not need to rely on the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e5dda242","line":119,"range":{"start_line":119,"start_character":16,"end_line":119,"end_character":21},"updated":"2019-01-10 16:36:12.000000000","message":"SR-IOV","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":129,"context_line":"Indirect Mode"},{"line_number":130,"context_line":"~~~~~~~~~~~~~"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"No other nova modifications are required for indirect mode SR-IOV."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Direct Mode"},{"line_number":135,"context_line":"~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_85ecd6f3","line":132,"range":{"start_line":132,"start_character":0,"end_line":132,"end_character":66},"updated":"2019-01-10 16:36:12.000000000","message":"On lines 63-68, you mention that we need to claim resources for indirect mode - so I guess \"no other\" here means \"besides resources claims, as described in the previous section\"","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":129,"context_line":"Indirect Mode"},{"line_number":130,"context_line":"~~~~~~~~~~~~~"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"No other nova modifications are required for indirect mode SR-IOV."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"Direct Mode"},{"line_number":135,"context_line":"~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_f0238004","line":132,"range":{"start_line":132,"start_character":0,"end_line":132,"end_character":66},"in_reply_to":"dfd5e7cf_85ecd6f3","updated":"2019-01-10 18:25:00.000000000","message":"yes ill state that explcitly","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":135,"context_line":"~~~~~~~~~~~"},{"line_number":136,"context_line":""},{"line_number":137,"context_line":"For direct mode SR-IOV, to enable live migration the SR-IOV devices must"},{"line_number":138,"context_line":"be first detached on the source after ``pre_live_migrate`` and then"},{"line_number":139,"context_line":"reattached in ``post_live_migration_at_destination``."},{"line_number":140,"context_line":""},{"line_number":141,"context_line":"This mimics the existing suspend [2]_ and resume [3]_ workflow where"},{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_25eacb74","line":139,"range":{"start_line":138,"start_character":9,"end_line":139,"end_character":53},"updated":"2019-01-10 15:53:46.000000000","message":"wouldn\u0027t this necessarily lead to network connectivity interruption? Isn\u0027t preventing network disruption one of the goals of live migration? i.e. the tenant shouldn\u0027t even notice their VM has moved hosts?\n\n[Later... OK, I see the .. note:: below on line 145]","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":135,"context_line":"~~~~~~~~~~~"},{"line_number":136,"context_line":""},{"line_number":137,"context_line":"For direct mode SR-IOV, to enable live migration the SR-IOV devices must"},{"line_number":138,"context_line":"be first detached on the source after ``pre_live_migrate`` and then"},{"line_number":139,"context_line":"reattached in ``post_live_migration_at_destination``."},{"line_number":140,"context_line":""},{"line_number":141,"context_line":"This mimics the existing suspend [2]_ and resume [3]_ workflow where"},{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_a87e8f16","line":139,"range":{"start_line":138,"start_character":9,"end_line":139,"end_character":53},"in_reply_to":"dfd5e7cf_25eacb74","updated":"2019-01-10 18:25:00.000000000","message":"right it would if you did not use bonding but the guest can mitigate this using a bond in the guest if the need.\n\nfor the vrouter case as peering between routers is often done over a different interface then the main dataplane workload you can negate the need for recreate the bgp peering relationships which can take 15-20mins to complete assuming the contol plane traffic uses a live migrateable port type e.g. a vswitch or macvtap port.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":138,"context_line":"be first detached on the source after ``pre_live_migrate`` and then"},{"line_number":139,"context_line":"reattached in ``post_live_migration_at_destination``."},{"line_number":140,"context_line":""},{"line_number":141,"context_line":"This mimics the existing suspend [2]_ and resume [3]_ workflow where"},{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":143,"context_line":"to disk operation."},{"line_number":144,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_aa5448cb","line":141,"range":{"start_line":141,"start_character":33,"end_line":141,"end_character":37},"updated":"2019-01-10 14:55:52.000000000","message":"like this :)","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":138,"context_line":"be first detached on the source after ``pre_live_migrate`` and then"},{"line_number":139,"context_line":"reattached in ``post_live_migration_at_destination``."},{"line_number":140,"context_line":""},{"line_number":141,"context_line":"This mimics the existing suspend [2]_ and resume [3]_ workflow where"},{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":143,"context_line":"to disk operation."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":".. note:: If you want to maintain network connectivity during the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e5036341","line":142,"range":{"start_line":141,"start_character":63,"end_line":142,"end_character":2},"updated":"2019-01-10 15:53:46.000000000","message":"whereby is a single word.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":138,"context_line":"be first detached on the source after ``pre_live_migrate`` and then"},{"line_number":139,"context_line":"reattached in ``post_live_migration_at_destination``."},{"line_number":140,"context_line":""},{"line_number":141,"context_line":"This mimics the existing suspend [2]_ and resume [3]_ workflow where"},{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":143,"context_line":"to disk operation."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":".. note:: If you want to maintain network connectivity during the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_3087c8da","line":142,"range":{"start_line":141,"start_character":63,"end_line":142,"end_character":2},"in_reply_to":"dfd5e7cf_e5036341","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"94f0a6f55cf0f97c53b296ccc294843fe171a2ba","unresolved":false,"context_lines":[{"line_number":144,"context_line":""},{"line_number":145,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":146,"context_line":"          migration, as the direct mode SR-IOV device will be detached,"},{"line_number":147,"context_line":"          a bond is required in the guest to a transparently live migratable"},{"line_number":148,"context_line":"          interface such as a vSwitch interface or a indirect mode SR-IOV"},{"line_number":149,"context_line":"          device. The recently added ``net_fallback`` kernel driver is out"},{"line_number":150,"context_line":"          of scope of this spec but could also be used."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_08a9dbdd","line":147,"range":{"start_line":147,"start_character":66,"end_line":147,"end_character":76},"updated":"2019-01-10 17:05:49.000000000","message":"nit: migrable ?","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":143,"context_line":"to disk operation."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":146,"context_line":"          migration, as the direct mode SR-IOV device will be detached,"},{"line_number":147,"context_line":"          a bond is required in the guest to a transparently live migratable"},{"line_number":148,"context_line":"          interface such as a vSwitch interface or a indirect mode SR-IOV"},{"line_number":149,"context_line":"          device. The recently added ``net_fallback`` kernel driver is out"},{"line_number":150,"context_line":"          of scope of this spec but could also be used."},{"line_number":151,"context_line":""},{"line_number":152,"context_line":""},{"line_number":153,"context_line":"XML Generation"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_eaf810c5","line":150,"range":{"start_line":145,"start_character":0,"end_line":150,"end_character":55},"updated":"2019-01-10 14:55:52.000000000","message":"I must admit I had expected to see something like this described here, given that detaching a network interface can be disruptive, but I guess that is additional work that requires co-operation from the guest. We can address in the future. Once question I do have though: if we detach and reattach, what does the guest see? Does the same interface reappear or is it a new interface? I ask because presumably the destination host could have different NICs (or the same NICs in different slots) which means something is likely to change. We don\u0027t need to do it here, but this must be documented once this is complete, IMO.\n\nLater: I see you documented this in the \"Other end user impact\" section. Nice job.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":142,"context_line":"by we workaround QEMUs inability to save device state during a suspend"},{"line_number":143,"context_line":"to disk operation."},{"line_number":144,"context_line":""},{"line_number":145,"context_line":".. note:: If you want to maintain network connectivity during the"},{"line_number":146,"context_line":"          migration, as the direct mode SR-IOV device will be detached,"},{"line_number":147,"context_line":"          a bond is required in the guest to a transparently live migratable"},{"line_number":148,"context_line":"          interface such as a vSwitch interface or a indirect mode SR-IOV"},{"line_number":149,"context_line":"          device. The recently added ``net_fallback`` kernel driver is out"},{"line_number":150,"context_line":"          of scope of this spec but could also be used."},{"line_number":151,"context_line":""},{"line_number":152,"context_line":""},{"line_number":153,"context_line":"XML Generation"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_10d084c9","line":150,"range":{"start_line":145,"start_character":0,"end_line":150,"end_character":55},"in_reply_to":"dfd5e7cf_eaf810c5","updated":"2019-01-10 18:25:00.000000000","message":"the guest will see a pcie or ahci hot unplug of the nic\nfollowed by a host plug of the nic on the destination which will prompt the kernel to fully reinitialize the nic.\n\nit could be a different nic yes but we do not have any way to specify the nic venor you desire today. traits or some other mechanism coudl be used in the future but today you can only say i want an sriov nic but not which one. that is out of scope of this spec but is something that would be nice to address at some point.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":168,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":169,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":170,"context_line":"  to fall back on cold migration. As this alternative would not fix the"},{"line_number":171,"context_line":"  problem of incomplete live migration supprot addtional documentation or"},{"line_number":172,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":173,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":174,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_ea21f04e","line":171,"range":{"start_line":171,"start_character":39,"end_line":171,"end_character":56},"updated":"2019-01-10 14:55:52.000000000","message":"support additional","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":168,"context_line":"* As always we could do nothing and continue to not support live migration"},{"line_number":169,"context_line":"  with SR-IOV devices. In this case, operators would have to continue"},{"line_number":170,"context_line":"  to fall back on cold migration. As this alternative would not fix the"},{"line_number":171,"context_line":"  problem of incomplete live migration supprot addtional documentation or"},{"line_number":172,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":173,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":174,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_10b164e2","line":171,"range":{"start_line":171,"start_character":39,"end_line":171,"end_character":56},"in_reply_to":"dfd5e7cf_ea21f04e","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":170,"context_line":"  to fall back on cold migration. As this alternative would not fix the"},{"line_number":171,"context_line":"  problem of incomplete live migration supprot addtional documentation or"},{"line_number":172,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":173,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":174,"context_line":""},{"line_number":175,"context_line":"* We could add a new API check to determine if an instance has an SR-IOV"},{"line_number":176,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_0ae23402","line":173,"range":{"start_line":173,"start_character":53,"end_line":173,"end_character":64},"updated":"2019-01-10 14:55:52.000000000","message":"like the recently introduced \"prevent live migration with NUMA topologies\" patch. I should probably write a backportable patch to do that for SR-IOV devices too...","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":170,"context_line":"  to fall back on cold migration. As this alternative would not fix the"},{"line_number":171,"context_line":"  problem of incomplete live migration supprot addtional documentation or"},{"line_number":172,"context_line":"  optionally a driver level check to reject live migration would be warranted"},{"line_number":173,"context_line":"  to protect operators that may not be aware of this limitation."},{"line_number":174,"context_line":""},{"line_number":175,"context_line":"* We could add a new API check to determine if an instance has an SR-IOV"},{"line_number":176,"context_line":"  port and explicitly fail to migrate in this case with a new error."}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_70bc70bb","line":173,"range":{"start_line":173,"start_character":53,"end_line":173,"end_character":64},"in_reply_to":"dfd5e7cf_0ae23402","updated":"2019-01-10 18:25:00.000000000","message":"ya it would be a similar safety net.","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":181,"context_line":""},{"line_number":182,"context_line":"It is expected that no data model changes should be required as the existing"},{"line_number":183,"context_line":"VIF object in the ``migration_data`` object should be able to store the"},{"line_number":184,"context_line":"associated PCI address info. if this is not the case a small extension to"},{"line_number":185,"context_line":"those objects will be required for this info."},{"line_number":186,"context_line":""},{"line_number":187,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_a5955a42","line":184,"range":{"start_line":184,"start_character":29,"end_line":184,"end_character":31},"updated":"2019-01-10 16:36:12.000000000","message":"If","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":181,"context_line":""},{"line_number":182,"context_line":"It is expected that no data model changes should be required as the existing"},{"line_number":183,"context_line":"VIF object in the ``migration_data`` object should be able to store the"},{"line_number":184,"context_line":"associated PCI address info. if this is not the case a small extension to"},{"line_number":185,"context_line":"those objects will be required for this info."},{"line_number":186,"context_line":""},{"line_number":187,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_d002fc3e","line":184,"range":{"start_line":184,"start_character":29,"end_line":184,"end_character":31},"in_reply_to":"dfd5e7cf_a5955a42","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":220,"context_line":""},{"line_number":221,"context_line":"SR-IOV live migration will be enabled if both the source and dest node support"},{"line_number":222,"context_line":"it. If either compute node does not support this feature the migration will"},{"line_number":223,"context_line":"be aboted by the conductor."},{"line_number":224,"context_line":""},{"line_number":225,"context_line":"Developer impact"},{"line_number":226,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_eacfb07c","line":223,"range":{"start_line":223,"start_character":3,"end_line":223,"end_character":9},"updated":"2019-01-10 14:55:52.000000000","message":"aborted","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":220,"context_line":""},{"line_number":221,"context_line":"SR-IOV live migration will be enabled if both the source and dest node support"},{"line_number":222,"context_line":"it. If either compute node does not support this feature the migration will"},{"line_number":223,"context_line":"be aboted by the conductor."},{"line_number":224,"context_line":""},{"line_number":225,"context_line":"Developer impact"},{"line_number":226,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_50dccc17","line":223,"range":{"start_line":223,"start_character":3,"end_line":223,"end_character":9},"in_reply_to":"dfd5e7cf_eacfb07c","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"6f35ec847821c87c2ca24b37ea3d95c79ea98044","unresolved":false,"context_lines":[{"line_number":234,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":235,"context_line":"however, as that will require both the source and dest node to"},{"line_number":236,"context_line":"support SR-IOV live migration first."},{"line_number":237,"context_line":"As a result this feature, will have no impact for this release."},{"line_number":238,"context_line":""},{"line_number":239,"context_line":"To ensure cross version compatiblity"},{"line_number":240,"context_line":"the conductor will validate if the source and destination nodes"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_05ad4730","line":237,"range":{"start_line":237,"start_character":11,"end_line":237,"end_character":24},"updated":"2019-01-10 15:53:46.000000000","message":"As a result, this feature will have...","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"16336b1b1b394a427ce52e830456ac191c440b3b","unresolved":false,"context_lines":[{"line_number":234,"context_line":"guests in the future by allowing live migration to be used"},{"line_number":235,"context_line":"however, as that will require both the source and dest node to"},{"line_number":236,"context_line":"support SR-IOV live migration first."},{"line_number":237,"context_line":"As a result this feature, will have no impact for this release."},{"line_number":238,"context_line":""},{"line_number":239,"context_line":"To ensure cross version compatiblity"},{"line_number":240,"context_line":"the conductor will validate if the source and destination nodes"}],"source_content_type":"text/x-rst","patch_set":10,"id":"bfdaf3ff_70e1d0d0","line":237,"range":{"start_line":237,"start_character":11,"end_line":237,"end_character":24},"in_reply_to":"dfd5e7cf_05ad4730","updated":"2019-01-10 18:25:00.000000000","message":"Done","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":278,"context_line":"Testing"},{"line_number":279,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":280,"context_line":""},{"line_number":281,"context_line":"This feature will be tested primarily via unit and functional test."},{"line_number":282,"context_line":"as SR-IOV testing is not available in the gate tempest test will not"},{"line_number":283,"context_line":"be possible. Third party ci could be implemented but that is not part"},{"line_number":284,"context_line":"of the scope of this spec. The use of the netdevsim kernel module to allow"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_45e10ed0","line":281,"range":{"start_line":281,"start_character":66,"end_line":281,"end_character":67},"updated":"2019-01-10 16:36:12.000000000","message":"comma here","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":280,"context_line":""},{"line_number":281,"context_line":"This feature will be tested primarily via unit and functional test."},{"line_number":282,"context_line":"as SR-IOV testing is not available in the gate tempest test will not"},{"line_number":283,"context_line":"be possible. Third party ci could be implemented but that is not part"},{"line_number":284,"context_line":"of the scope of this spec. The use of the netdevsim kernel module to allow"},{"line_number":285,"context_line":"testing of SR-IOV without SR-IOV hardware was evaluated. While the netdevsim"},{"line_number":286,"context_line":"kernel module does allow the creation of an SR-IOV PF netdev and the"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_e5e7e2e3","line":283,"range":{"start_line":283,"start_character":25,"end_line":283,"end_character":27},"updated":"2019-01-10 16:36:12.000000000","message":"CI","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"2a3c6c635fccde719c6a11f11e22619808d47dd8","unresolved":false,"context_lines":[{"line_number":284,"context_line":"of the scope of this spec. The use of the netdevsim kernel module to allow"},{"line_number":285,"context_line":"testing of SR-IOV without SR-IOV hardware was evaluated. While the netdevsim"},{"line_number":286,"context_line":"kernel module does allow the creation of an SR-IOV PF netdev and the"},{"line_number":287,"context_line":"allocation of SR-IOV VF netdevs, it does not simulate pci devices."},{"line_number":288,"context_line":"As a result in its current form the netdevsim kernel module cannot be used"},{"line_number":289,"context_line":"to enable SR-IOV testing in the gate."},{"line_number":290,"context_line":""}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_a5d1dafd","line":287,"range":{"start_line":287,"start_character":54,"end_line":287,"end_character":57},"updated":"2019-01-10 16:36:12.000000000","message":"PCI","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"35eb7c4d82c10b22ce5c91b2d3be26be3b767baa","unresolved":false,"context_lines":[{"line_number":281,"context_line":"This feature will be tested primarily via unit and functional test."},{"line_number":282,"context_line":"as SR-IOV testing is not available in the gate tempest test will not"},{"line_number":283,"context_line":"be possible. Third party ci could be implemented but that is not part"},{"line_number":284,"context_line":"of the scope of this spec. The use of the netdevsim kernel module to allow"},{"line_number":285,"context_line":"testing of SR-IOV without SR-IOV hardware was evaluated. While the netdevsim"},{"line_number":286,"context_line":"kernel module does allow the creation of an SR-IOV PF netdev and the"},{"line_number":287,"context_line":"allocation of SR-IOV VF netdevs, it does not simulate pci devices."},{"line_number":288,"context_line":"As a result in its current form the netdevsim kernel module cannot be used"},{"line_number":289,"context_line":"to enable SR-IOV testing in the gate."},{"line_number":290,"context_line":""},{"line_number":291,"context_line":"Documentation Impact"},{"line_number":292,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":10,"id":"dfd5e7cf_2aab7889","line":289,"range":{"start_line":284,"start_character":27,"end_line":289,"end_character":37},"updated":"2019-01-10 14:55:52.000000000","message":"As before, ++ for including this here","commit_id":"82533c6b32302b1e32f7204e70c88b49d9ba4b38"},{"author":{"_account_id":8864,"name":"Artom Lifshitz","email":"notartom@gmail.com","username":"artom"},"change_message_id":"a15b88ff3ae590be02124905904c998c1a6a9463","unresolved":false,"context_lines":[{"line_number":130,"context_line":"~~~~~~~~~~~~~"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"No other nova modifications are required for indirect mode SR-IOV"},{"line_number":133,"context_line":"beyond those already covered in the Resouce claims sechtion."},{"line_number":134,"context_line":""},{"line_number":135,"context_line":"Direct Mode"},{"line_number":136,"context_line":"~~~~~~~~~~~"}],"source_content_type":"text/x-rst","patch_set":13,"id":"bfdaf3ff_16563c08","line":133,"range":{"start_line":133,"start_character":51,"end_line":133,"end_character":59},"updated":"2019-01-10 19:46:21.000000000","message":"Dude.","commit_id":"1096a447af7eb4f94cceb1d13402068b6b1d8f4d"}]}
