)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"16a98200ee97671b1ab3cc6de3f32df8448b2175","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"fa1a9e52_ae5a7fc3","updated":"2024-06-11 14:32:50.000000000","message":"@Brian, what do you think about this patch in general? I wonder if we should make it a feature flag and let it to validated in the field before making any change in defaults. I don\u0027t think we will be able to test it in gate properly, even if I manage to set up raft job (lots of work for that.) I may try to squeeze this patch into perf testing downstream, but even then - some real world deployments may have even larger scale to be sure that the switch to non-leaders won\u0027t affect anyone negatively.","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"},{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"e9c0babc1724bdc6c38dc292e5036de323f4b5bf","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"2aa628b6_e8dc21e0","updated":"2024-10-14 16:08:04.000000000","message":"FYI I\u0027m still waiting for my scale team to report back about their testing for this patch. I also understand that Brian points out to a consideration from OVS code that we should be aware. I am not sure how to reconcile this yet with the recommendation that came from Ilya (OVS core working on db layer) to switch to this mode. Will need to follow up with Ilya on this one again.","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"},{"author":{"_account_id":1131,"name":"Brian Haley","email":"haleyb.dev@gmail.com","username":"brian-haley"},"change_message_id":"5f49cd256793c94f9cddb1fa4a4f4d048f2396a8","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"827a9723_79ab5bf4","updated":"2024-08-08 17:15:47.000000000","message":"Hi Ihar - I generally like it, and think if we put it behind a config option it would be useful. I\u0027m going to ping our OVS team and get a second opinion, and point them at Ilya\u0027s email as well.","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"},{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"7b48cf42722992d3abd0a58875c8b9a0cb3e2181","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"12e9568b_a65d940f","updated":"2024-06-07 01:24:08.000000000","message":"Looking at all multinode jobs, I don\u0027t think there\u0027s any job in any of our pipelines here that would run raft cluster. So this change cannot be properly validated in the openstack gate. We\u0027ll have to test it on some other premises that would allow to spin up a properly sized cluster with at least 3 controller nodes.","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"},{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"269475dced1de60609c193da0ca11370d03509ad","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"a13b20b2_58050873","updated":"2024-06-06 20:01:03.000000000","message":"check experimental","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"},{"author":{"_account_id":9656,"name":"Ihar Hrachyshka","email":"ihrachys@redhat.com","username":"ihrachys","status":"Red Hat Networking Systems Engineer"},"change_message_id":"a2ada89ce4c600b5e28e76f749e4dd2ef9d8df11","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":2,"id":"27ba0fdb_440e9ba1","updated":"2024-11-05 23:13:13.000000000","message":"Brian\u0027s concern is still not addressed; rebased to keep it without conflicts.","commit_id":"f168a6a02f3a37d30d86fea203002e00226feebd"},{"author":{"_account_id":5756,"name":"Terry Wilson","email":"twilson@redhat.com","username":"otherwiseguy"},"change_message_id":"59c3993a0d6032c31f2d8f70e5997f77c105f78b","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":6,"id":"ad646489_0602b9bf","updated":"2025-08-18 13:30:06.000000000","message":"I haven\u0027t read the code yet, but we can\u0027t run *all* as leader_only\u003dfalse. Anything that uses ovsdb locks must me leader_only\u003dTrue (e.g. MaintenanceWorker) because OVSDB locks are per-server and not per cluster. So, using raft, it will behave as always unlocked 2/3 of the time.","commit_id":"034c297c1fe1baaaad572c73742ddba293c11e32"},{"author":{"_account_id":1131,"name":"Brian Haley","email":"haleyb.dev@gmail.com","username":"brian-haley"},"change_message_id":"1e8319d32595e0093a9826edea7531e594260743","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":6,"id":"3bd0122c_83c2f746","updated":"2025-08-17 21:36:32.000000000","message":"check experimental","commit_id":"034c297c1fe1baaaad572c73742ddba293c11e32"}],"neutron/plugins/ml2/drivers/ovn/mech_driver/ovsdb/ovsdb_monitor.py":[{"author":{"_account_id":1131,"name":"Brian Haley","email":"haleyb.dev@gmail.com","username":"brian-haley"},"change_message_id":"abb7d66eb85df50b722a9912dd46a15c40916238","unresolved":true,"context_lines":[{"line_number":739,"context_line":"        try:"},{"line_number":740,"context_line":"            super(Ml2OvnIdlBase, self).__init__("},{"line_number":741,"context_line":"                remote, schema, probe_interval\u003dprobe_interval,"},{"line_number":742,"context_line":"                leader_only\u003dFalse, **kwargs)"},{"line_number":743,"context_line":"        except TypeError:"},{"line_number":744,"context_line":"            # TODO(twilson) We can remove this when we require ovs\u003e\u003d2.12.0"},{"line_number":745,"context_line":"            super(Ml2OvnIdlBase, self).__init__("}],"source_content_type":"text/x-python","patch_set":1,"id":"b86ce512_b3e83974","line":742,"updated":"2024-08-27 22:01:47.000000000","message":"So looking at the OVS DB code, the comment regarding leader_only bothers me a little, specifically the part about a read acting on stale data:\n\n/* By default, or if \u0027leader_only\u0027 is true, when \u0027idl\u0027 connects to a clustered\n * database, the IDL will avoid servers other than the cluster leader. This\n * ensures that any data that it reads and reports is up-to-date.  If\n * \u0027leader_only\u0027 is false, the IDL will accept any server in the cluster, which\n * means that for read-only transactions it can report and act on stale data\n * (transactions that modify the database are always serialized even with false\n * \u0027leader_only\u0027).  Refer to Understanding Cluster Consistency in ovsdb(7) for\n * more information. */\n \n I\u0027m not sure what implications that has for neutron-server.","commit_id":"f54f899cd33c17c2e32becfa8f80f842745c3b20"}]}