Earlier  
Posted Nick Remark
#openstack-nova - 2021-06-10
12:24:34 opendevreview sean mooney proposed openstack/os-vif master: [WIP] mock ovs.poller.Poller https://review.opendev.org/c/openstack/os-vif/+/795770
12:25:14 sean-k-mooney gibi: that passes the os-vif functional tests which actully create ports in ovs and locally i had it assert that the Poller was called
12:25:27 sean-k-mooney but im not sure if this will actully fix the issue
12:25:56 sean-k-mooney ill take a look at the tempest run when its done and we can see if the repeating dbug message is still present or not
12:29:21 kashyap lyarwood: </me back after eclipse hunting> Hey. Have you got a bug filed, or shall I file one?
12:29:21 gibi sean-k-mooney: cool
12:29:41 kashyap [OT] A couple of pictures of projections, if you missed it: https://kashyapc.fedorapeople.org/partial_solar_eclipse_2021/
12:29:47 lyarwood kashyap: I don't yet so feel free to write one up if you have time
12:29:59 kashyap lyarwood: Sure; I'll do it right now
12:30:16 kashyap But would be good to see if we can reproduce this at least twice...
12:32:09 gibi kashyap: I think we hit it 7 times in the last 7 days http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22libvirt.libvirtError%3A%20unable%20to%20connect%20to%20server%20at%5C%22
12:32:57 kashyap gibi: Ah, that's good to know.
12:33:13 kashyap So the trigger is block-migrating a puased instance
12:33:35 kashyap (Where "block-migrating" == migrating the instance along with its storage)
13:21:38 lyarwood kashyap: https://bugs.launchpad.net/nova/+bug/1912310 found an older bug we could mark as a duplicate if you've already written a fresh bug up
13:21:38 opendevmeet Launchpad bug 1912310 in OpenStack Compute (nova) "libvirt.libvirtError: unable to connect to server at " [Medium,Confirmed]
13:22:35 kashyap lyarwood: Ah, good find; I'm just drafting it locally
13:22:43 lyarwood https://bugs.launchpad.net/nova/+bug/1797185 is another
13:22:43 opendevmeet Launchpad bug 1797185 in OpenStack Compute (nova) "live migration intermittently fails in CI with "Connection refused" during guest transfer" [Low,Confirmed]
13:22:48 kashyap Sigh
13:23:33 kashyap lyarwood: What I wonder is - how can we check _why_ the destination is unreahable...
13:23:41 kashyap Is it something peculiar to our upstream CI; or something else
13:26:10 lyarwood I honestly think it's a bug in the libvirt python bindings where they miss that the migration has already failed, try to poll it's progress and don't handle the fact that it's no longer there correctly
13:26:23 lyarwood the unreachable error is a red herring IMHO
13:27:47 kashyap lyarwood: Yeah; I asked Michal from libvirt; and he suggested Jiri Denemark
13:28:07 kashyap He isn't around; but I'll ask him to comment on this once he's back
13:28:24 kashyap lyarwood: Also, you only speak of Python bindings - why won't it be a bug in the C API itself?
13:28:37 kashyap Asking out of ignorance, not challenging :)
13:28:49 bauzas hola
13:29:00 bauzas lyarwood: gibican we know how many jobs have problems ?
13:29:04 bauzas gibi: ^
13:29:35 bauzas if we have problems with fixing those problems, can we make those specific jobs non-voting until we fix the main issues ?
13:29:39 lyarwood kashyap: it could be that as well but n-cpu is calling the python bindings and they are raising the error here so it easily be an issue there
13:30:20 bauzas fwiw, I saw that https://review.opendev.org/c/openstack/nova/+/791506 passed the check pipeline but we don't know yet whether the gate one will accept it
13:30:34 kashyap lyarwood: Ah, okay; you're just speaking of the error trace in our case. Reasonable
13:31:08 kashyap lyarwood: Can you copy/paste your comment from the review to start with in there?
13:31:18 kashyap (In the latest bug you filed 2021 Jan)
13:31:30 lyarwood bauzas: There are various intermittent failures at the moment, I'm trying to clean up the gate-failures bug tag to get a better handle on what is failing and how often
13:32:02 lyarwood bauzas: we could move them to non-voting but this has been going on for weeks so I'd rather we try to get a better handle on this first
13:32:04 bauzas okay, fair enough
13:32:27 lyarwood I honestly think we haven't been landing enough to notice recently
13:32:28 bauzas lyarwood: I guess my point is that we're still having *all* the changes getting -1 because of the mypy issue
13:32:51 bauzas so I want the gate back asap
13:33:06 lyarwood yup fair
13:33:07 bauzas we have two possibilities, the main fix https://review.opendev.org/c/openstack/nova/+/791506
13:33:38 bauzas but I also prepared an alternative change that's simplier for the gate and doesn't hit the transient issues we know https://review.opendev.org/c/openstack/nova/+/795744
13:34:17 bauzas so, I guess, I'd recommend to wait for the main change to merge, but if we get a Zuul -2 from the gate on it, I'd honestly recommend to let https://review.opendev.org/c/openstack/nova/+/795744 go
13:42:23 lyarwood sean-k-mooney: https://bugs.launchpad.net/nova/+bug/1863889/comments/3 another possible hit of the ovsdbapp issue
13:42:23 opendevmeet Launchpad bug 1863889 in OpenStack Compute (nova) "Revert resize problem in neutron-tempest-dvr-ha-multinode-full" [Medium,Confirmed]
13:44:06 sean-k-mooney lyarwood: im talking to otherwiseguy and ralonsoh in #openstack-neutron about it now but ill let them know
13:44:28 lyarwood it's just another random timeout FWIW
13:44:44 lyarwood with lots of spam from ovsdbapp so no hard proof
13:55:31 bauzas lyarwood: the problem is that logstash isn't reliable
13:55:44 bauzas despite some people like me and dansmith expressing our needs
13:56:07 bauzas we need to focus on the few things we saw occuring often
13:56:20 bauzas and either fix them or hide them under the carpet until we are in a better situation
14:01:25 lyarwood yup agreed
14:01:49 lyarwood and that's what I have been doing prior to now with things like https://bugs.launchpad.net/nova/+bug/1929710 for example
14:01:49 opendevmeet Launchpad bug 1929710 in OpenStack Compute (nova) "virDomainGetBlockJobInfo fails during swap_volume as disk '$disk' not found in domain" [Medium,New]
14:03:08 lyarwood https://bugs.launchpad.net/nova/+bugs?field.tag=gate-failure should be in a much better state now, I've moved loads of things to incomplete where I couldn't find any evidence of recent hits etc.
14:03:23 lyarwood these should close automatically in the next 60 days
14:08:31 bauzas lyarwood: ++ thanks for triaging
14:22:05 opendevreview Merged openstack/nova-specs master: Speed up server details https://review.opendev.org/c/openstack/nova-specs/+/791620
14:36:57 kashyap lyarwood: Before I forget, for now, I've copy/pasted (with attribution) your comment from the change here: https://bugs.launchpad.net/nova/+bug/1912310
14:36:57 opendevmeet Launchpad bug 1912310 in OpenStack Compute (nova) "libvirt.libvirtError: unable to connect to server at " [Medium,Confirmed]
14:40:12 gibi lyarwood: thanks for the cleanup. I add that query to the weekly agenda
14:40:48 lyarwood kashyap: ack thanks
14:40:53 lyarwood gibi: awesome cheers
14:54:23 sean-k-mooney gibi: lyarwood mocking the poller wont work but i have a seperate patch to move it to a real thread against ovsdb and talking to otherwiseguy https://bugs.launchpad.net/neutron/+bug/1930926 and https://review.opendev.org/c/openstack/neutron/+/794892 might also help if we ported the same change to os-vif
14:54:23 opendevmeet Launchpad bug 1930926 in neutron "Failing over OVN dbs can cause original controller to permanently lose connection" [Medium,Fix released] - Assigned to Terry Wilson (otherwiseguy)
14:54:38 sean-k-mooney mocking the poller actully cause the delay
14:54:59 sean-k-mooney as it forces use to reconnect every 5 seconds sicne we dont respond to the server side echo
14:56:05 lyarwood `# Overwriting globals in a library is clearly a good idea` lol
14:56:49 lyarwood sean-k-mooney: ack kk
14:58:54 sean-k-mooney ya...
15:00:04 gibi sean-k-mooney: will look at it shortly
15:32:25 bauzas mmm, /me needs to leave, unfortunately too early to see outcomes of https://zuul.opendev.org/t/openstack/status#795533
15:32:51 bauzas hopefully the change will be merged, but the failing jobs are currently still running
15:54:17 gibi sean-k-mooney: is the the new directon? https://review.opendev.org/c/openstack/ovsdbapp/+/795789
16:10:01 opendevreview Balazs Gibizer proposed openstack/nova master: Test the NotificationFixture https://review.opendev.org/c/openstack/nova/+/758450
16:10:01 opendevreview Balazs Gibizer proposed openstack/nova master: Move fake_notifier impl under NotificationFixture https://review.opendev.org/c/openstack/nova/+/758451
16:10:02 opendevreview Balazs Gibizer proposed openstack/nova master: rpc: Mark attributes as private https://review.opendev.org/c/openstack/nova/+/792803
16:11:09 sean-k-mooney gibi: its potentally part of it where are some other optimistaion that they have done in neutron that we could port ot os-vif that might help
16:11:23 stephenfin grenade-multinode failed on the deps change again :( tempest.api.compute.admin.test_live_migration.LiveMigrationTest.test_live_block_migration_with_attached_volume this time
16:11:37 gibi stephenfin: :/
16:11:40 sean-k-mooney gibi: it sound like the issue really need to be fixed in the ovs python bindindings
16:12:00 gibi sean-k-mooney: so no easy fix?
16:12:20 gibi sean-k-mooney: should we switch back to vsctl in the gate until we make a fix for the bindig?
16:12:34 gibi stephenfin: looking
16:12:37 sean-k-mooney gibi: my next attempt will be to patch get_system_poll in the ovs binding
16:12:46 gibi sean-k-mooney: OK, lets try that
16:12:46 sean-k-mooney to not return the unpatch version form os-vif
16:13:14 sean-k-mooney gibi: the short term fix would be to revert to the deprecated driver in devstack
16:13:29 sean-k-mooney untill we figure out a way to work around this properly
16:13:51 sean-k-mooney so we can certenly do that if we want too
16:13:54 gibi sean-k-mooney: lets keep the devstack option open, but I think it is OK to try patching get_system_poll first
16:14:10 gibi let see if that helps, if not then go with the devstack change
16:14:40 sean-k-mooney basicaly if i have get_system_poll returng _SelectSelect then i think it shoudl not block
16:14:43 sean-k-mooney on reconnect
16:15:20 sean-k-mooney then the neutron chagnes to disbale the echo and instead use tcp keepalive and limit the tables we cache woudl also miniums reconnect time

Earlier   Later