)]}'
{"os_traits/compute/__init__.py":[{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"33492021684fa1a9935a983e87dfeb7d3e8ed1b6","unresolved":false,"context_lines":[{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the "},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_28ed2a23","line":32,"range":{"start_line":32,"start_character":69,"end_line":32,"end_character":70},"updated":"2019-12-13 17:13:49.000000000","message":"whitespace damage","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"d93b9100cdefaa2e7a20bb461dd37a17f4f87789","unresolved":false,"context_lines":[{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the "},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_a8703a0a","line":32,"range":{"start_line":32,"start_character":69,"end_line":32,"end_character":70},"in_reply_to":"3fa7e38b_28ed2a23","updated":"2019-12-13 17:36:16.000000000","message":"Done","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"33492021684fa1a9935a983e87dfeb7d3e8ed1b6","unresolved":false,"context_lines":[{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027COMPUTE_ACCELERATORS_BUILD\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_68e7a247","line":40,"range":{"start_line":40,"start_character":5,"end_line":40,"end_character":12},"updated":"2019-12-13 17:13:49.000000000","message":"I don\u0027t think the COMPUTE_ prefix matches the other things here.\n\nFrom the looks of it, just \"ACCELERATORS\" would be appropriate?","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"ba297219a4d854476ebb22c03a6bba9301389314","unresolved":false,"context_lines":[{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027COMPUTE_ACCELERATORS_BUILD\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_488186f9","line":40,"range":{"start_line":40,"start_character":5,"end_line":40,"end_character":12},"in_reply_to":"3fa7e38b_28220a55","updated":"2019-12-13 17:27:39.000000000","message":"There is no support for anything without the ability during build, which is why I didn\u0027t include build in my example. See \"trusted certs\" or \"device tagging\" for example.\n\nFor the hot attach case, we\u0027ll definitely need RPC changes, and adding that won\u0027t affect schedule-ability. So I\u0027d expect to handle the upgrade impact of that with RPC and service versions and thus we\u0027d never have another trait specifically for that.","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"17d8e030096c5b391e01f213808cd89c27ee073c","unresolved":false,"context_lines":[{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027COMPUTE_ACCELERATORS_BUILD\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_c8e816a8","line":40,"range":{"start_line":40,"start_character":5,"end_line":40,"end_character":12},"in_reply_to":"3fa7e38b_68e7a247","updated":"2019-12-13 17:33:30.000000000","message":"\u003e I don\u0027t think the COMPUTE_ prefix matches the other things here.\n \u003e \n \u003e From the looks of it, just \"ACCELERATORS\" would be appropriate?\n\nCorrect; as written this will result in a trait called COMPUTE_COMPUTE_ACCELERATORS.","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"04841a5cecef3fddc80351802c9cc7684858daec","unresolved":false,"context_lines":[{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027COMPUTE_ACCELERATORS_BUILD\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_28220a55","line":40,"range":{"start_line":40,"start_character":5,"end_line":40,"end_character":12},"in_reply_to":"3fa7e38b_68e7a247","updated":"2019-12-13 17:20:34.000000000","message":"We want to distinguish between the ability to build a VM and a future ability to attach accelerators to a VM. So, ACCELERATORS_BUILD (vs. possible ACCELERATORS_ATTACH later)?","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"d93b9100cdefaa2e7a20bb461dd37a17f4f87789","unresolved":false,"context_lines":[{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027COMPUTE_ACCELERATORS_BUILD\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":1,"id":"3fa7e38b_c86b3619","line":40,"range":{"start_line":40,"start_character":5,"end_line":40,"end_character":12},"in_reply_to":"3fa7e38b_c8e816a8","updated":"2019-12-13 17:36:16.000000000","message":"Done","commit_id":"e09db27791367392bc9af5e89ef3b4ea18015d2a"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"48a1f7bdb33c363cda484ca6650a3be82da46f32","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_881bfe7d","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"updated":"2019-12-13 17:49:30.000000000","message":"This is really too much implementation detail, specific to how we plan to use this right now. I won\u0027t insist you take it out -- but what is missing is a basic description of what the trait itself is/means. Something like:\n\n # Indicates that the compute manager and driver are capable of processing\n # requests for accelerators in concert with the accelerator service","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"39244605cce884542ead1395ae4df1627cd0699d","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_67610c39","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"in_reply_to":"3fa7e38b_24dc7244","updated":"2019-12-13 21:47:04.000000000","message":"I think it helps to provide context on why a trait was added and how it is expected to be used. But, frankly, I am fine either way. Removing it.","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"d95232b829460dce4e920edb4114bd7e673f48c9","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_8aed2fba","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"in_reply_to":"3fa7e38b_67610c39","updated":"2019-12-13 22:43:09.000000000","message":"\u003e I think it helps to provide context on why a trait was added and\n \u003e how it is expected to be used.\n\nYup, the copy of this information in the commit message is perfect for that purpose.","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"ad8c64310024d73fd37f29d562bdd1aaeb5ab0df","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_c32a3776","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"in_reply_to":"3fa7e38b_881bfe7d","updated":"2019-12-13 17:52:04.000000000","message":"Is that not what I am saying with: \"The compute manager supports handling accelerator requests.\"?","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"dcad9dffb22f7ff0d29b20d24c60d921e4cb02c1","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_24dc7244","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"in_reply_to":"3fa7e38b_9e281498","updated":"2019-12-13 20:15:55.000000000","message":"I think we should nuke L32-39. How we use it is not something that should be documented here.","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"715e3b3429e9f7669bf8f6a4f039330b1a2355bf","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    # When a Cyborg device profile name is present in the flavor, the"},{"line_number":33,"context_line":"    # conductor will initiate creation and binding of accelerator requests,"},{"line_number":34,"context_line":"    # and the compute manager must wait for the ARQ binding notifications."},{"line_number":35,"context_line":"    # In a deployment where the conductor is new, but one or more compute"},{"line_number":36,"context_line":"    # managers are older than the needed version, this flow will break."},{"line_number":37,"context_line":"    # To prevent that: (a) the compute manager publishes this trait,"},{"line_number":38,"context_line":"    # and (b) the scheduler factors this trait in the Placement query, so"},{"line_number":39,"context_line":"    # that older compute nodes are not included in allocation candidates."},{"line_number":40,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":41,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":2,"id":"3fa7e38b_9e281498","line":39,"range":{"start_line":31,"start_character":0,"end_line":39,"end_character":73},"in_reply_to":"3fa7e38b_c32a3776","updated":"2019-12-13 19:45:10.000000000","message":"Heh, yes it is, my bad for not reading closely enough.\n\nStill think the rest is superfluous / overly specific. But fine with leaving.","commit_id":"285c9c40211cdd97b3869493b2227905cf1ce9ee"},{"author":{"_account_id":14070,"name":"Eric Fried","email":"openstack@fried.cc","username":"efried"},"change_message_id":"d95232b829460dce4e920edb4114bd7e673f48c9","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":33,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":3,"id":"3fa7e38b_6a2a33e7","line":31,"range":{"start_line":31,"start_character":10,"end_line":31,"end_character":25},"updated":"2019-12-13 22:43:09.000000000","message":"technically \"virt driver\"","commit_id":"b8a92d689ee808ca2b875d3d2be8ec4dc27419e5"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"4ab4c88fe029bef5fcc155e9c6e54a4b22aef2eb","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":33,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":3,"id":"3fa7e38b_7c1a64c7","line":31,"range":{"start_line":31,"start_character":10,"end_line":31,"end_character":25},"in_reply_to":"3fa7e38b_40330414","updated":"2020-01-03 07:10:22.000000000","message":"Agree with both of you, however the compute manager needs to have the right code to handle accelerator request events [1].\n\n[1] https://review.opendev.org/#/c/631244/53/nova/compute/manager.py@2719","commit_id":"b8a92d689ee808ca2b875d3d2be8ec4dc27419e5"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"c4879d1c7ae490e76fed5a3fb7802a202713c9e2","unresolved":false,"context_lines":[{"line_number":28,"context_line":"    # not make sense to use in GET /allocation_candidates until/unless there is"},{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":33,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":3,"id":"3fa7e38b_40330414","line":31,"range":{"start_line":31,"start_character":10,"end_line":31,"end_character":25},"in_reply_to":"3fa7e38b_6a2a33e7","updated":"2020-01-02 15:14:34.000000000","message":"yes technically its the virt driver that reports support but based on you +2 im reading this as a nit rather then an objection.","commit_id":"b8a92d689ee808ca2b875d3d2be8ec4dc27419e5"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"c4879d1c7ae490e76fed5a3fb7802a202713c9e2","unresolved":false,"context_lines":[{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":33,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":3,"id":"3fa7e38b_fb6d01cb","line":32,"range":{"start_line":32,"start_character":4,"end_line":32,"end_character":18},"updated":"2020-01-02 15:14:34.000000000","message":"nit: nova technically already supprots acclerators in several virt drivers via VGPU support in libvirt or pci passthough support in general. so part of me whish we would use cycborg\nor ACCELERATOR_REQUESTS here instead but but that is just a nit so im ok with ACCELERATORS even if it would not be my first choice.","commit_id":"b8a92d689ee808ca2b875d3d2be8ec4dc27419e5"},{"author":{"_account_id":21672,"name":"Sundar Nadathur","email":"sundar.nadathur@intel.com","username":"nsundar"},"change_message_id":"4ab4c88fe029bef5fcc155e9c6e54a4b22aef2eb","unresolved":false,"context_lines":[{"line_number":29,"context_line":"    # a way to tie it to the condition of SAME_HOST-ness."},{"line_number":30,"context_line":"    \u0027SAME_HOST_COLD_MIGRATE\u0027,"},{"line_number":31,"context_line":"    # The compute manager supports handling accelerator requests."},{"line_number":32,"context_line":"    \u0027ACCELERATORS\u0027,"},{"line_number":33,"context_line":"]"}],"source_content_type":"text/x-python","patch_set":3,"id":"3fa7e38b_3c762c0e","line":32,"range":{"start_line":32,"start_character":4,"end_line":32,"end_character":18},"in_reply_to":"3fa7e38b_fb6d01cb","updated":"2020-01-03 07:10:22.000000000","message":"Understand your concern, however I suspect most name choices can be contentious, so I appreciate that you are on board.","commit_id":"b8a92d689ee808ca2b875d3d2be8ec4dc27419e5"}]}