)]}'
{"nova/conf/workarounds.py":[{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"ec79890a2e86ab1cc5d4df217167771dd1abb6f9","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_2db04528","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"updated":"2018-12-18 14:20:41.000000000","message":"if we back port this which im ok with \nwe should default to True on the sable branches.\n\nthe reasoning behind this is reletivly simply.\n\n1.) this would be a behavior change to backprot with the default of false as live migration however broken would no longer \"work\" following a minor update.\n\n2.) ovs-dpdk, vpp, snabb, and vrouter all support vhost-user which requires hugepages to be used as nova cannot configure a non hugepage memory backing currently that meets there needs. one of the key factors in calculating the appropriateness of any of these vhost user backends over sriov is the fact they allow live migration. defaulting to False would retroactively change that calculation. if yo remove livemigraton from equation the only motivating reasons for not using sriov are security groups and vxlan/tunneling support.\n\nnote that vhost-user backends are most common in NFV deployments where hardware tends to be uniform. additionally\nwhen huge pages are used without CPU pinning the failure rate\nfor live migration is much lower and in production is typically not an issue.\n\n3.) this can cause issue when people use cpu pinning or have\nnon uniform hosts so this is valuable to backport for deployment that are not using vhost-user backends. as such i am not arguing we should not backport the feature but just the default on stable is wrong.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":5754,"name":"Alex Xu","email":"hejie.xu@intel.com","username":"xuhj"},"change_message_id":"4fee87219b2ee14fb43358eb09855b13e2865080","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"ffd0ebdf_0059f958","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_1339a39f","updated":"2019-01-04 08:04:24.000000000","message":"breaking the VM with live migration is wrose than nfv user can\u0027t live migration, and nfv user can change this configure, then enable the live migration again. so...I choose to set as False.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":15334,"name":"Stephen Finucane","display_name":"stephenfin","email":"stephenfin@redhat.com","username":"sfinucan"},"change_message_id":"ee57777ca5750d7088019bf73a784ebe2ac4469a","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_1339a39f","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_20d6caf9","updated":"2018-12-21 09:54:30.000000000","message":"\u003e Regardless, deployment tools are going to want/need to expose a\n \u003e knob to tweak this and document it so the downstream impacts of\n \u003e this are non-zero.\n\nFor what it\u0027s worth, we are tracking this internally from the TripleO side. I don\u0027t know if Kolla properly supports these NFV\u0027y things yet (the session I was in in Berlin suggests not) and I can\u0027t speak for the others.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":8768,"name":"Chris Friesen","email":"chris.friesen@windriver.com","username":"cbf123"},"change_message_id":"06dae36fc32342a329620074a3f303603c1125ab","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_7335bad2","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_2db04528","updated":"2018-12-18 15:46:16.000000000","message":"I can see where you\u0027re coming from Sean, but I think we should be safe by default, which would suggest defaulting to False.\n\nYes, users with NFV deployments will likely want this to be True, but those users are also probably advanced enough to be able to set their config options as part of the upgrade process rather than just blindly using the defaults.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":12898,"name":"Tony Breeds","email":"tony@bakeyournoodle.com","username":"tonyb"},"change_message_id":"23d1a5afcec10141fdd54757c115cb667ec8a0b3","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_20d6caf9","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_36871b0d","updated":"2018-12-21 05:25:13.000000000","message":"So we can either introduce a behaviour change on a stable branch (and be \u0027safe by default\u0027) or we can keep the broken behaviour.\n\n Rock | This review | Hard-place\n\nIf we were faced with this dilemma in the context of a CVE, we\u0027d pick \u0027safe by default\u0027 accept the behaviour change release a minor update and probably wouldn\u0027t loose any sleep.  There isn\u0027t a CVE here (that we know of) but I think the scope of the bug is large enough to warrant the behaviour change.\n\nRegardless, deployment tools are going to want/need to expose a knob to tweak this and document it so the downstream impacts of this are non-zero.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"6aaa41a269c3a48fc007b279e8a9216753cd23af","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_fe02f170","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_7335bad2","updated":"2018-12-18 16:45:38.000000000","message":"I tend to agree with Chris - this is really tied to a bug because admins could unknowingly be breaking VMs by live migrating them. We have an upgrade release note and we could release this as a minor version (it\u0027s not a feature, it\u0027s a precaution). But I\u0027ve added Tony to see what he thinks.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":11604,"name":"sean mooney","email":"smooney@redhat.com","username":"sean-k-mooney"},"change_message_id":"3b9e5f35ded28b8abc018d5f33fd27a54caba63b","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_36871b0d","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_b0a336fd","updated":"2018-12-19 13:32:25.000000000","message":"so i strongly disagree with backporting a behavioral change if we dont allow exsiting usecase to work without config/flvor changes but if  im being over ruled i can live with the cahnge even if i dont like it.\n\ni agree nfv deployment will be advanced enough to know how to set this and i belive they are the group most likely to want to change this. to that point ill push to have our deployment tooling set this to true on all host with dpdk networking on upgrade.\n\nfor new deployments defaulting to safe is resonable but we should break exsting deployments on upgrade and for dpdk deployment where live migration is expected to work and be supported defaulting to false in our upgrade tooling would be a major regression. it should be simple to simple to change the default to true just for the dpdk deployments so we likely don\u0027t need to have it default to true in the code.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"},{"author":{"_account_id":10135,"name":"Lee Yarwood","display_name":"Lee Yarwood","email":"lyarwood@redhat.com","username":"lyarwood"},"change_message_id":"103320532431fa985f5d9d3aebedc71996b71719","unresolved":false,"context_lines":[{"line_number":221,"context_line":""},{"line_number":222,"context_line":"    cfg.BoolOpt("},{"line_number":223,"context_line":"        \u0027enable_numa_live_migration\u0027,"},{"line_number":224,"context_line":"        default\u003dFalse,"},{"line_number":225,"context_line":"        help\u003d\"\"\""},{"line_number":226,"context_line":"Enable live migration of instances with NUMA topologies."},{"line_number":227,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_b0a336fd","line":224,"range":{"start_line":224,"start_character":8,"end_line":224,"end_character":21},"in_reply_to":"3f79a3b5_fe02f170","updated":"2018-12-19 08:36:00.000000000","message":"FWIW I also agree with Chris and Matt here. I\u0027d rather operators opt-in to this broken behaviour even if it was previously the default on the stable branches. If anything this should increase their awareness of the pending solution referenced in the release note.","commit_id":"52b89734426253f64b6d4797ba4d849c3020fb52"}]}