)]}'
{"specs/ocata/approved/libvirt-driver-for-veritas-hyperscale.rst":[{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"d41de823e15ab6940df50dcc9c432efdbf198790","unresolved":false,"context_lines":[{"line_number":154,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Cinder blueprint for Veritas HyperScale"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Testing"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba5da102_2b2c7c01","line":157,"updated":"2016-11-01 08:52:35.000000000","message":"The main dependancy is support in libvirt and QEMU.  NEither have a release which includes the code yet. The QEMU 2.8 freeze date is today and the vxhs driver is not merged, so it seems unlikely to be available before QEMU 2.9, which is too late to consider this for Nova in Ocata.","commit_id":"5581abda076cde8f95626c7945f3746fa34e8d09"},{"author":{"_account_id":23610,"name":"Ketan Nilangekar","email":"ketonne@yahoo.com","username":"ketan.nilangekar"},"change_message_id":"9b1ea4fe0d3b66c0b5a25290a621e7b49ccfea3c","unresolved":false,"context_lines":[{"line_number":154,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Cinder blueprint for Veritas HyperScale"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Testing"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba5da102_ee4a7925","line":157,"in_reply_to":"ba5da102_2b2c7c01","updated":"2016-11-01 20:11:48.000000000","message":"QEMU/libvirt is one of the options that we have for IO transfer. We also have capability of using an nfs based IO tap in HyperScale.\nThe current nova implementation itself does not have a hard dependency on QEMU/libvirt or any other IO transfer mechanisms.\nWe do have multiple use cases for vxhs requiring high throughput as well as ones which are ok with NFS level performance.","commit_id":"5581abda076cde8f95626c7945f3746fa34e8d09"},{"author":{"_account_id":23610,"name":"Ketan Nilangekar","email":"ketonne@yahoo.com","username":"ketan.nilangekar"},"change_message_id":"c2718cc1c1384b7e4a834a8301bf3f859b9f9f1e","unresolved":false,"context_lines":[{"line_number":154,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Cinder blueprint for Veritas HyperScale"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Testing"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba5da102_6dec7a2a","line":157,"in_reply_to":"ba5da102_cd312d5c","updated":"2016-11-14 05:21:27.000000000","message":"NFS based IO tap is not a short term workaround. In fact it is a viable option for HyperScale for certain lower end workloads. We have no issues supporting multiple IO taps and our storage back-end is designed to ingest IO that way. In future we may also support NBD, iSCSI or other technologies depending on the hypervisors that we work with. We are certainly looking to support non-KVM hypervisors in the future.\nAlso the nova driver extension should not/does not have any dependencies on the IO tap that we are using.\nI will update the spec to reflect this soon.","commit_id":"5581abda076cde8f95626c7945f3746fa34e8d09"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"28a3af52e68cbd356211829db46b7ee543cc38d4","unresolved":false,"context_lines":[{"line_number":154,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"Cinder blueprint for Veritas HyperScale"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"Testing"},{"line_number":159,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":160,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"ba5da102_cd312d5c","line":157,"in_reply_to":"ba5da102_ee4a7925","updated":"2016-11-02 09:08:35.000000000","message":"The text in the \"proposed change\" section above does not mention anything about using NFS instead of the QEMU driver. I\u0027m not in favour of adding support to NFS as a short term workaround to deal with fact that QEMU support is not yet finished. That means we\u0027ll end up having to support two distinct parallel I/O stacks for the same technology in order to preserve back-compat for people who initially deployed. So IMHO we should wait for the QEMU work to be completed and use that exclusively.","commit_id":"5581abda076cde8f95626c7945f3746fa34e8d09"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"207d6e5e5d95561cefa411584be7f307504860c2","unresolved":false,"context_lines":[{"line_number":67,"context_line":"QEMU block driver that needs to be installed on compute nodes."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"This method of tapping IO is being upstreamed to QEMU and tracked here:"},{"line_number":70,"context_line":"https://bugzilla.redhat.com/show_bug.cgi?id\u003d1341866"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Once this is done, and this driver is installed, nova nodes"},{"line_number":73,"context_line":"can be provisioned to use Veritas HyperScale storage system and use them just"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7a77a97e_e0504b0d","line":70,"updated":"2016-11-21 10:31:56.000000000","message":"That work has missed QEMU 2.8 release, so is not likely to be available until QEMU 2.9 release. That will be too late to consider this for the Nova Ocata release.","commit_id":"b58f45d4a09d64485dbe00c0e1342806edc0d718"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"ee29796514cca99255113c9b15ef1bffeca34fec","unresolved":false,"context_lines":[{"line_number":67,"context_line":"QEMU block driver that needs to be installed on compute nodes."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"This method of tapping IO is being upstreamed to QEMU and tracked here:"},{"line_number":70,"context_line":"https://bugzilla.redhat.com/show_bug.cgi?id\u003d1341866"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Once this is done, and this driver is installed, nova nodes"},{"line_number":73,"context_line":"can be provisioned to use Veritas HyperScale storage system and use them just"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7a77a97e_0298e769","line":70,"in_reply_to":"7a77a97e_02562706","updated":"2016-11-21 12:53:48.000000000","message":"The way you\u0027ve described it suggests you\u0027re targeting the QEMU impl first, and the NFS/iSCSI stuff later, whereas it looks like you\u0027ll need the exact opposite if you\u0027re to do this in Ocata.","commit_id":"b58f45d4a09d64485dbe00c0e1342806edc0d718"},{"author":{"_account_id":23610,"name":"Ketan Nilangekar","email":"ketonne@yahoo.com","username":"ketan.nilangekar"},"change_message_id":"b2135c91e21c51beeb2f0f40495cf76e7ff05572","unresolved":false,"context_lines":[{"line_number":67,"context_line":"QEMU block driver that needs to be installed on compute nodes."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"This method of tapping IO is being upstreamed to QEMU and tracked here:"},{"line_number":70,"context_line":"https://bugzilla.redhat.com/show_bug.cgi?id\u003d1341866"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Once this is done, and this driver is installed, nova nodes"},{"line_number":73,"context_line":"can be provisioned to use Veritas HyperScale storage system and use them just"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7a77a97e_04d47445","line":70,"in_reply_to":"7a77a97e_0298e769","updated":"2016-11-21 13:08:59.000000000","message":"We already have an NFS implementation whose orchestration is abstracted as a series of cli calls. The cli will also support QEMU virtual disk provisioning. The nova driver will be agnostic of which method is being used for IO tap.\nThere will be no references to either QEMU or NFS in the driver. Based on the use case, the Veritas HyperScale installation will configure the right kind of IO tap.","commit_id":"b58f45d4a09d64485dbe00c0e1342806edc0d718"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"f520e45692df0bf21085731e8d17196967e7fbad","unresolved":false,"context_lines":[{"line_number":67,"context_line":"QEMU block driver that needs to be installed on compute nodes."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"This method of tapping IO is being upstreamed to QEMU and tracked here:"},{"line_number":70,"context_line":"https://bugzilla.redhat.com/show_bug.cgi?id\u003d1341866"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Once this is done, and this driver is installed, nova nodes"},{"line_number":73,"context_line":"can be provisioned to use Veritas HyperScale storage system and use them just"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a74a57a_8d305e03","line":70,"in_reply_to":"7a77a97e_04d47445","updated":"2016-11-24 15:01:50.000000000","message":"What you are describing is *not* possible. Having seen the proposed QEMU patches for VXHS integration, it is clear that Nova will have to configure libvirt and QEMU differently depending on whether its using the native QEMU driver for vxhs or NFS. That configuration can\u0027t be magically hidden by the external veritas tools","commit_id":"b58f45d4a09d64485dbe00c0e1342806edc0d718"},{"author":{"_account_id":23610,"name":"Ketan Nilangekar","email":"ketonne@yahoo.com","username":"ketan.nilangekar"},"change_message_id":"f83172f1ad1e89637d42548f5e6d39965d94caa7","unresolved":false,"context_lines":[{"line_number":67,"context_line":"QEMU block driver that needs to be installed on compute nodes."},{"line_number":68,"context_line":""},{"line_number":69,"context_line":"This method of tapping IO is being upstreamed to QEMU and tracked here:"},{"line_number":70,"context_line":"https://bugzilla.redhat.com/show_bug.cgi?id\u003d1341866"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"Once this is done, and this driver is installed, nova nodes"},{"line_number":73,"context_line":"can be provisioned to use Veritas HyperScale storage system and use them just"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7a77a97e_02562706","line":70,"in_reply_to":"7a77a97e_e0504b0d","updated":"2016-11-21 12:42:45.000000000","message":"The spec is only mentioning QEMU as one of the IO tapping mechanisms. The driver implementation will sufficiently abstract the IO tap and will not be dependent on any specific IO tap.\nThe spec clearly mentions this in the last paragraph of this section.\nIf you feel the spec needs more clarity regarding this please let me know.","commit_id":"b58f45d4a09d64485dbe00c0e1342806edc0d718"}]}