Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2026-08-17
16:38:57 lajoskatona https://review.opendev.org/c/openstack/nova/+/982085 (Suppress error in cleanup of pre live migration )
16:39:31 lajoskatona and the other one is for handling queue_size/mtu related values: https://review.opendev.org/c/openstack/nova/+/995841/3
16:39:40 lajoskatona thanks in advance
16:39:41 Uggla lajoskatona 👍, I will add to the list if not already done.
16:39:55 gmaan one thing on gate status, i observed nova-grenade-multinode job failing for test_shelve_unshelve_server test more frequently last week
16:40:02 gmaan i reported bug for that
16:40:04 gmaan #link https://bugs.launchpad.net/nova/+bug/2163552
16:40:15 lajoskatona Uggla: thanks, please let me know if I need for some extra meeting or similar
16:40:26 gmaan i did not debug it yet but in case anyone see similar failure, this is known one
16:40:54 Uggla gmann, I have seen it. I mean the bug.
16:41:26 gmaan k, I will try to debug it after FF and happen more frequently
16:41:52 Uggla gmaan any idea of the frequency ?
16:41:53 gmaan at least 4-5 times I saw this in different changes
16:42:01 gmaan #link https://zuul.opendev.org/t/openstack/builds?job_name=nova-grenade-multinode&skip=0
16:42:09 gmaan this ^^ is state currenyty whihc is not bad
16:42:30 sean-k-mooney lajoskatona: updatign the mtu durring live migratrion is not a thing we supprot
16:42:54 sean-k-mooney lajoskatona: and there is a good reason fro that which si we do not allow the mtu to be chagne on a port that is atched to a vm
16:43:03 sean-k-mooney i kno nuetron can allow that but nova has never supproted that
16:43:18 lajoskatona sean-k-mooney: it 's not updating the value, but not changing if on the dst there is a different thing in the cfg, and breaking things in the VM
16:43:20 sean-k-mooney also adding or removign the mtu elemetn changes the virtio options
16:43:26 sean-k-mooney which make the vm un live mgiratable
16:43:35 sean-k-mooney so updating it woudl only work if the element already exitsed
16:44:18 sean-k-mooney lajoskatona: so the value of the mtu in the xml shoudl not change if the dest port bining has a diffent value
16:44:41 sean-k-mooney i.e. if you enable jumb frames on a network
16:44:51 sean-k-mooney that shoudl only reuslt in teh xml value updatign after a hard reboot
16:44:55 sean-k-mooney not on a live migration
16:46:09 lajoskatona the patch handles only the case when the nova-comp.cfg is different, nothing from Neutron side today (life is difficult even without that) and the same can happen with min/max_queue_size as the bug you opened (https://bugs.launchpad.net/nova/+bug/1854844)
16:46:40 sean-k-mooney we only use the mtu values form teh nova config as a fallback
16:46:47 sean-k-mooney and that is really only there for nova-networks
16:46:58 sean-k-mooney so the real fix is to stop usign those values entirly
16:47:09 sean-k-mooney or even remove them at this point
16:47:24 lajoskatona ok, that can be also a good point to avoid confusion with config
16:47:53 sean-k-mooney for the quese leght and count that is difent
16:47:58 sean-k-mooney as that is entrily config driven
16:48:01 lajoskatona so than the MTU part can be forgotten now (and let's check if the cfg can be deprecated and removed)
16:48:11 sean-k-mooney but for mtu you shoudl not config that in nova if your using neutron
16:48:33 sean-k-mooney ya for the queue config those cant change
16:48:40 sean-k-mooney and we need to use the souce value on live migriton
16:48:55 sean-k-mooney as that changes the way the nic is virutalsied to the guest
16:49:10 sean-k-mooney i think qemu will not alow the queue count to chagne whiel the guest is running
16:49:40 lajoskatona actually I got it from field where they upgraded from old nova/libvirt and during migrating around the VMs the MTU value mismatch had broken their apps, but if that is not used any more let's remove it
16:50:13 lajoskatona but I don't want to keep the meeting open, we can discuss this topic offline, or in gerrit
16:50:36 Uggla lajoskatona 👍
16:50:42 Uggla anything else ? If not I think we can close and keep this meeting.
16:51:33 sean-k-mooney lajoskatona: it was common when changing form ml2/ovs to ml2/ovn
16:51:33 bryanfraschetti Hi folks, I'm hoping to get some reviews on this series: https://review.opendev.org/q/topic:%22improve_attachment_resiliance_against_timeout%22
16:51:40 sean-k-mooney but the fix was to cold migrate
16:51:50 sean-k-mooney not to hack live meigtion to chnge the mtu
16:52:05 sean-k-mooney lajoskatona: we can chat about this more later
16:52:30 lajoskatona sean-k-mooney: ack, thanks
16:52:56 Uggla bryanfraschetti, at least I will put your series in the list.
16:53:28 bryanfraschetti Thank you!
16:54:10 Uggla anything else ?
16:55:01 Uggla No so Closing, thanks for joining this meeting. Have a nice day/evening.
16:55:11 Uggla Thanks for joining this meeting. Have a nice day/evening.
16:55:17 opendevmeet Log: https://meetings.opendev.org/meetings/nova/2026/nova.2026-08-17-16.00.log.html
16:55:17 opendevmeet Minutes (text): https://meetings.opendev.org/meetings/nova/2026/nova.2026-08-17-16.00.txt
16:55:17 opendevmeet Minutes: https://meetings.opendev.org/meetings/nova/2026/nova.2026-08-17-16.00.html
16:55:17 opendevmeet Meeting ended Mon Aug 17 16:55:17 2026 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
16:55:17 Uggla #endmeeting
16:55:23 lajoskatona o/
17:30:59 dansmith gmaan: I'm not sure I understand this: https://github.com/openstack/nova/blob/master/nova/compute/rpcapi.py#L981
17:31:56 dansmith this lets us start new live migrations after shutdown is already going.. I think your comment is about conductor having already checked with both source and dest, but is there any reason we need to actually proceed during shutdown there?
17:32:01 dansmith like some state conductor has altered?
17:32:25 dansmith point being, you can continue issuing new live migration requests during a shutdown which will eventually run out the timeout clock and cause a failure
17:33:42 gmaan dansmith: no, new live migratin will be rejected by check_can_live_migrate when conductor check it on source and dest. check_can_live_migrate_ is listneing on rpcserver which is stopped durign shutdown
17:34:17 dansmith ah okay, cool
17:34:32 gmaan that is why check_can_live_migrate make sure both source and dest are not shutting down so proceed rest of the live migration on rpc server alt in cae shutdown happen in betwen
17:34:42 dansmith gotcha
17:56:51 gouthamr dansmith: hey there, ack on the virtiofs cold-migration notes from the nova meeting, and ty for explaining stuff on the patch. i think i agree with your hardening suggestion more than any pointed fix i can make to the recovery paths.. i'm testing an update
17:58:26 gouthamr dansmith: ack also on the PTO :) i'm out on and for a couple days after feature freeze day too (27th Aug) .. but i'll react to any review feedback as soon as i can
17:59:01 gouthamr and i'll bother melwitt, Uggla and anyone else that was looking at this :) ty!
18:26:06 opendevreview Ghanshyam Maan proposed openstack/nova master: Use service shutdown related config options in nova-graceful-shutdown https://review.opendev.org/c/openstack/nova/+/1001017
18:38:32 opendevreview Ghanshyam Maan proposed openstack/nova master: Improve the graceful shutdown job ansible tasks name https://review.opendev.org/c/openstack/nova/+/1001212
19:04:14 opendevreview Alexandre Engloner proposed openstack/nova master: Typo fix in List Hypervisors details Policy path. https://review.opendev.org/c/openstack/nova/+/1001216
20:01:43 opendevreview Clif Houck proposed openstack/nova master: Parallelize per-node resource updates https://review.opendev.org/c/openstack/nova/+/980679
20:01:43 opendevreview Clif Houck proposed openstack/nova master: perf(ironic): eliminate O(N²) ProviderTree deepcopy at startup https://review.opendev.org/c/openstack/nova/+/980676
20:35:14 opendevreview Merged openstack/nova master: Stabilize func test waiting for VM ERROR state https://review.opendev.org/c/openstack/nova/+/998822
22:45:11 opendevreview Goutham Pacha Ravi proposed openstack/nova master: Support cold migration with virtiofs shares https://review.opendev.org/c/openstack/nova/+/989633
#openstack-nova - 2026-08-18
01:02:21 opendevreview Goutham Pacha Ravi proposed openstack/nova master: Support cold migration with virtiofs shares https://review.opendev.org/c/openstack/nova/+/989633
01:59:12 opendevreview Merged openstack/nova master: Stabilize functional test for slow instance.save https://review.opendev.org/c/openstack/nova/+/999121
06:15:53 opendevreview Goutham Pacha Ravi proposed openstack/nova master: Support cold migration with virtiofs shares https://review.opendev.org/c/openstack/nova/+/989633
09:09:50 gibi gmaan: I did some digging into the logs of https://bugs.launchpad.net/nova/+bug/2163552 it is strange as everything seems to correct, except that nova-api does not forwards the vif-plugged event to the compute even though the compute already waiting for it.
09:14:03 gibi gmaan: funnily that sounds very similar to the downstream bug report you triaged recently https://redhat.atlassian.net/browse/OSPRH-31354
09:14:46 gibi and that is reported to Antelope
09:14:52 gibi and we see it on master
09:14:55 gibi crazy
09:44:02 opendevreview Balazs Gibizer proposed openstack/nova master: [DNM]: Troubleshoot race condition in unshelve vif plug https://review.opendev.org/c/openstack/nova/+/1001294
09:44:19 gibi gmaan: ^^ I added extra logs
11:33:08 opendevreview Balazs Gibizer proposed openstack/nova master: [DNM]: Troubleshoot race condition in unshelve vif plug https://review.opendev.org/c/openstack/nova/+/1001294
12:45:28 opendevreview Merged openstack/nova master: Make functional-py313-threading voting https://review.opendev.org/c/openstack/nova/+/997216
13:35:17 dansmith gibi: what is the purpose of the StaticallyDelayingCancellableTaskExecutorWrapper ?
13:36:34 dansmith oh for filtering out stops when a start is right on the heels
13:47:48 opendevreview Merged openstack/nova master: libvirt: Check only x86_64 arch for SEV capability detection https://review.opendev.org/c/openstack/nova/+/998285
13:51:27 gibi dansmith: yeah that is an old behavior of the libvirt virt driver
13:51:40 gibi just eventlet had a native submit_after facility to handle that
13:51:42 dansmith I remember discussing it now
13:51:55 dansmith it's a very (impressively) long class name
13:52:22 gibi I'm open to suggestions :) (we can still change futurist before the deadline);5;13~
13:53:37 dansmith are you building that into futurist?

Earlier   Later