Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2022-08-31
14:55:00 dansmith so actually
14:55:09 dansmith I think gibi might be right and that the feature itself is broken
14:55:27 dansmith but I can't explain why it's working on single-cell, since we're targeted to cell0 by default
14:55:35 dansmith it looks like we don't target until we hit superconductor
14:55:48 dansmith https://github.com/openstack/nova/blob/cbc9b516fb5e6b079b57cc40380e1d730751cd6c/nova/conductor/manager.py#L974-L975
14:56:22 dansmith so I can't explain why this ever works, because the api should have only checked for the host being present in cell0, which should never have any hosts
14:56:33 gibi I think I see where we target the context
14:56:34 gibi _get_instance_from_cell
14:56:37 gibi https://github.com/openstack/nova/blob/3862cfc649d0099971c91d17e6f8800d10712a26/nova/compute/api.py#L2867-L2870
14:56:44 gibi this actually has a side effect to alter the context
14:57:15 dansmith ah, yep, okay I missed that
14:57:19 dansmith phew :)
14:57:20 gibi so unshelve actually works within a cell
14:57:39 gibi we need to document that it does not work across cells and it need to emit a better error
14:57:50 dansmith yeah
14:58:09 sean-k-mooney ya although that proably a latent bug
14:58:20 sean-k-mooney i.e. if we use the old way of targeting an az
14:58:26 sean-k-mooney and that az was in a differnt cell
14:58:32 sean-k-mooney it likely would have failed in teh same place
14:58:42 dansmith latent bug in terms of raising a useful error you mean?
14:58:47 sean-k-mooney yes
14:58:50 dansmith yeah
14:58:59 sean-k-mooney its just now that we have a tempest test that hits it we are seeing it
14:59:17 sean-k-mooney but we proably were not doing cross az unshelve in the multi cell job
15:00:47 gibi I've update the bug
15:00:56 dansmith sean-k-mooney: can you +W the skip so it's headed in after tests?
15:01:10 dansmith or gmann
15:01:54 gibi Uggla: probably the above discussion about unshelve is interesting to you too ;)
15:01:58 gibi Uggla: also https://bugs.launchpad.net/nova/+bug/1988316 :)
15:03:31 gmann dansmith: gibi done
15:03:36 gibi gmann: thanks
15:03:45 dansmith sweet
15:04:05 gmann I think its worth to add multi-cell jo on tempest side too and check the new tests for cross-cell thigns
15:04:09 gmann job
15:04:16 gmann I will do
15:04:44 gibi nah then we had the necessary gate block at FF week done
15:06:02 gibi could have been worse
15:06:31 gmann ? did not get
15:06:48 gmann you mean we should not add in tempest gate ?
15:06:59 gibi nah I'm just joking
15:07:00 dansmith I believe gibi is being funny :)
15:07:08 gmann :)
15:07:18 gibi so we tend to break the gate during FF as we are rushing
15:07:19 dansmith but in all seriousness, gmann adding multicell to tempest gate might be a bit heavy and nova-centric
15:07:30 dansmith not sure it's really necessary
15:07:36 dansmith maybe in experimental or something
15:07:47 gmann yeah experimental is good idea
15:09:23 bauzas add this in periodic
15:09:28 bauzas periodic-weelky
15:09:39 bauzas and then I'll check its state every week before the meeting
15:09:56 sean-k-mooney dansmith: sorry was geting a drink yes i can
15:09:58 bauzas better than experimental where not all of us would check it
15:10:08 sean-k-mooney oh gmann already did
15:11:24 gmann bauzas: with experimental we can check it during the new test addition. it run in nova gate as voting so periodic in tempest side would not be much benefit. if anything broken then nova gate will check it nd inform tempest
15:11:38 sean-k-mooney im not sure periodic-weekly helps
15:11:50 sean-k-mooney we will notice it on the nova side before it runs
15:11:54 gmann yeah
15:11:59 bauzas okok
15:12:07 bauzas maybe I misunderstood
15:12:50 dansmith right, periodic won't really help
15:12:59 gmann bauzas: we are talking to add multi-cell job in tempest experimental pipeline so that we can check new tests (migration/unshelve one ) for cross cells env
15:13:16 bauzas aaaaah
15:13:23 bauzas OK, then yes
15:13:29 bauzas fine with me
15:13:38 bauzas I was thinking it was a nova job
15:13:58 bauzas lgmt then to be experimental
16:29:37 gmann bauzas: with the new change in RBAC (dropped system scope from policy), we do not need this blueprint change anymore. I have added the notes there https://blueprints.launchpad.net/nova/+spec/allow-project-admin-list-hypervisors
16:30:16 gmann bauzas: please close/mark it not needed
16:32:33 gmann added the same notes in etherpad too for your easy tracking https://etherpad.opendev.org/p/nova-zed-blueprint-status
16:33:44 sean-k-mooney gmann i just set it to obsolete
16:34:34 gmann sean-k-mooney: thanks
16:58:22 bauzas ricolin: just said -1 for https://review.opendev.org/c/openstack/nova/+/830646 but if you don't have time today or tomorrow morning, lemme know
16:58:36 bauzas ricolin: we could just need a follow-up patch for my nits
16:58:47 bauzas and I'll +2
17:15:40 ricolin bauzas: thanks, I will try to update it ASAP
17:16:19 whoami-rajat dansmith, do you have any idea about the nova multi cell breaking constantly or is it just flaky?
17:16:29 dansmith whoami-rajat: there's a fix in the queue
17:16:35 dansmith whoami-rajat: no point in rechecking until that lands
17:16:48 whoami-rajat ah ok
17:16:51 whoami-rajat wasn't aware
17:16:51 dansmith it has been in check for hours, not even in the gate yet
17:17:09 whoami-rajat doesn't sound good
17:30:36 dansmith whoami-rajat: it's in the gate now
17:30:48 whoami-rajat great
17:30:58 whoami-rajat can i have the link?
17:31:15 dansmith it's 855378
17:31:24 whoami-rajat ack
17:56:48 sean-k-mooney whoami-rajat: sicne your change is approved even if its not merged by COB tommrow at m3 it can be rechecked to land it
17:56:59 sean-k-mooney so we will keep an eye on it until its landed
17:57:45 sean-k-mooney whoami-rajat: if you have a depends on form the nova client and osc patchs against the top patch we can review those in parralel
17:57:55 whoami-rajat sean-k-mooney, thanks, just wanted to update other project patches after this merged, guess it's safe to update the clients and tempest in parallel so will get to that
17:58:06 whoami-rajat sean-k-mooney, yep, i will just update them quickly and let you know
17:58:47 sean-k-mooney yep depends on will prevent them merging until the api change lands and then they can just be rechecked at that point
18:00:01 whoami-rajat ack, sounds good
18:14:14 opendevreview Rico Lin proposed openstack/nova master: libvirt: Add vIOMMU device to guest https://review.opendev.org/c/openstack/nova/+/830646
18:14:15 opendevreview Rico Lin proposed openstack/nova master: Add traits for viommu model https://review.opendev.org/c/openstack/nova/+/844507
18:15:12 sean-k-mooney ricolin: you still have not adress my feedback on https://review.opendev.org/c/openstack/nova/+/844507/15
18:15:45 sean-k-mooney the way you are doing the traits reporting is wrong
18:16:54 opendevreview Rajat Dhasmana proposed openstack/python-novaclient master: Add support to rebuild boot volume 2.93 https://review.opendev.org/c/openstack/python-novaclient/+/827163
18:18:08 whoami-rajat novaclient: https://review.opendev.org/c/openstack/python-novaclient/+/827163

Earlier   Later