Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2025-02-25
11:43:48 sean-k-mooney mikal: client freeze is gerneally the saem as FF
11:45:09 sean-k-mooney frickler: the 22.04 hwe kernel is basiclly the kernels form the 3 non lts release until the next lts. 24.04 baseline might be similar but 24.04 hwe will be newer
11:46:23 sean-k-mooney frickler: its not critical but i was hoping that it might help with some of the kernel panic we see although that genearly seam to be related to swtiching form the initramfs so it might not actuly be a "kernel" bug so much as an issue with how we are booting
11:57:38 sean-k-mooney frickler: i like cirros and dfisk image builder (i know cirros does not use that) but on think i hav ebeen meening to look into is bootc images i.e. bootable iamge build with podman and https://github.com/osbuild/bootc-image-builder fedora-minimal is 48MB i wonder how much bigger that woudl be if i added the kernel via the bootc build process
11:58:50 sean-k-mooney i got pretty far with my alpine poc a few years ago but never had time to complete it
11:59:46 sean-k-mooney friday is a hack/hussle day technailly at redhat so i might actuly use it to hack on somethign related to cirros or another minimal image
12:00:13 sean-k-mooney that or ill get distracted and work on feature freeze stuff :)
12:01:32 sean-k-mooney the normal fedora bootc image is like 900 MB which defaute the goal fo a small test image because there are plent fo cloud image out there that are smaller or the same size
12:48:59 opendevreview Douglas Viroel proposed openstack/nova master: Add support for showing scheduler_hints in server details https://review.opendev.org/c/openstack/nova/+/938604
14:34:08 opendevreview ribaudr proposed openstack/nova master: Update driver to deal with managed flag https://review.opendev.org/c/openstack/nova/+/938405
14:34:08 opendevreview ribaudr proposed openstack/nova master: Add managed flag to PCI device specification https://review.opendev.org/c/openstack/nova/+/937649
14:54:43 opendevreview ribaudr proposed openstack/nova master: Update driver to deal with managed flag https://review.opendev.org/c/openstack/nova/+/938405
14:54:43 opendevreview ribaudr proposed openstack/nova master: Add managed flag to PCI device specification https://review.opendev.org/c/openstack/nova/+/937649
15:25:48 sean-k-mooney Uggla: can you confirm the tags are proply updated on the comptue node tabel too
15:25:59 sean-k-mooney for https://review.opendev.org/c/openstack/nova/+/937649
15:29:13 sean-k-mooney Uggla: we do not require it for the manage feature but the tags are includied in the pci_stats column in the ComputeNode table https://github.com/openstack/nova/blob/master/nova/db/main/models.py#L227-L229
15:29:29 sean-k-mooney we will need that for the live migratable tag
15:30:01 sean-k-mooney i expect the existing logig to proprly update that without you needing to do anything
15:30:05 sean-k-mooney for managed
15:30:12 sean-k-mooney but it woudl be good to confirm
15:31:03 gibi sean-k-mooney: do you mean if the live-migratable flag is removed from the dev_spec then the pci pool containing the matching device should also be updated?
15:31:08 Uggla sean-k-mooney, you mean if this is register in the DB table ? --> https://paste.opendev.org/show/ba4BTb1bDKjCU052R3UP/
15:31:28 gibi I think that is fairly implict as the pools are memory only and the tag is removed with a nova-compute service restart
15:31:35 sean-k-mooney the schduler does nto use the pci_devices table
15:31:45 sean-k-mooney it uses the pci_stats form the compute node table
15:31:53 gibi ahh
15:31:57 gibi then one more table to look at
15:32:01 sean-k-mooney so the tag for "live migratable" will be stroed in the pci_stats
15:32:18 gibi I assumed pci stats in memory only and not persisted, but then I was wrong
15:32:21 sean-k-mooney gibi: think of pools as placement inventoies
15:32:21 gibi TIL
15:32:36 sean-k-mooney they just have the count and tags instead of traits
15:32:58 sean-k-mooney ya no this is needed for multi cell
15:33:14 sean-k-mooney we cant pass the statc via the scheduler update
15:33:18 sean-k-mooney rpc
15:33:31 sean-k-mooney the scdhluer need to be able to load them form the cell db when creatign the host state object
15:35:22 sean-k-mooney gibi: we dont need to look at managed in the schduer so this is really only relevnet ot the second series but the behvior shoudld be the same. the pools shoudl be in sync with the config
15:35:45 gibi ack
15:36:30 sean-k-mooney https://github.com/openstack/nova/blob/master/nova/pci/stats.py#L66-L79
15:36:41 sean-k-mooney mangage may be something we should add to the ignored tags
15:38:16 sean-k-mooney i can leave a comment to that effect on the reviews but i just notice the patches have been updated
15:38:55 sean-k-mooney i have not reviewd the most recent version yet. wehn i saw the paste only had the pci device table in it i was wondering if this had been missed
15:39:02 sean-k-mooney but i have not looked at the code to confirm
15:39:51 gibi sean-k-mooney: I think (weakly) that having extra tags on the device / pool is not an issue as the pool matching logic should be looking at the requested tags from the request side and ignore any extra on the pool side
15:40:34 gibi this is the a response for the ignored tags comment
15:40:39 sean-k-mooney correct and the ovo defines the tags as a dict of nullable strings
15:40:41 sean-k-mooney https://github.com/openstack/nova/blob/master/nova/objects/pci_device_pool.py#L36
15:40:54 sean-k-mooney so its not an object change
15:41:21 sean-k-mooney however we should keep the summery in the pools small in general to keep the memory overhead in the schduler and rpc size down
15:42:08 sean-k-mooney i think this could be a follow up patch but we likely shoudl filter out managed since we do not have a usecase for it in the schduler currently since managed is not in the alias
15:42:08 gibi wait do we send the Pool across the wire? you just said above scheduler loads it from the DB
15:42:39 sean-k-mooney i belive we send it across the wire in a few cases.
15:42:54 sean-k-mooney we have the schduler resouce view update perodic
15:43:18 sean-k-mooney where the compute manger updates the schduler with info about the instance on the host ectra
15:43:54 sean-k-mooney but its also going to get pulled when we do a hypervior list. which i guess is not over rpc
15:43:58 sean-k-mooney but directly from the db
15:44:19 gibi yeah hypervisor list only goes to RPC if it wants to get the uptime from the compute
15:44:27 gibi anyhow
15:44:37 sean-k-mooney i think that was zen only?
15:44:51 sean-k-mooney also deleteed in newer api versions so im not that worries about that
15:45:19 gibi so for the FF we need to check if pci_stats table is also clean after a tag removal. The ignored_tags thing might wait as it does not seem to break anything
15:46:03 sean-k-mooney yep
15:46:17 sean-k-mooney it would only be in the pool if you opted into the feature anyway
15:46:45 sean-k-mooney for live migration its imporant that that flag is there so we can select a device that supports it in teh pci passthough filter
15:47:59 gibi I fail to see the pci_stats table in my deployment....
15:48:15 sean-k-mooney its not a table
15:48:23 sean-k-mooney its a colum on the compute_node table
15:48:31 gibi aah
15:48:32 gibi OK
15:48:55 sean-k-mooney so its pulled anytime we pull a compute node object unless we are explcitly lazyloading that
15:49:33 sean-k-mooney its saved back every time update aviable resocues commits back the updated values
15:57:14 sean-k-mooney so its updated here https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py#L1377-L1382 and then its converted into the db format on compute node save by https://github.com/openstack/nova/blob/master/nova/objects/compute_node.py#L321-L326
16:00:40 opendevmeet The meeting name has been set to 'nova'
16:00:40 opendevmeet Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
16:00:40 opendevmeet Meeting started Tue Feb 25 16:00:40 2025 UTC and is due to finish in 60 minutes. The chair is bauzas. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:40 bauzas #startmeeting nova
16:01:40 gibi sean-k-mooney: https://paste.opendev.org/show/bEWhDvmH5NPNozp8V4pU/ confirmed the pci_stats also cleaned
16:02:12 sean-k-mooney gibi: cool i expected it to be. its just it should not be there at all. but we can adress that in a trivial followup
16:02:20 sean-k-mooney or dicsuss it more after the meeting
16:02:36 bauzas #link https://wiki.openstack.org/wiki/Meetings/Nova#Agenda_for_next_meeting
16:02:42 elodilles o/
16:02:44 bauzas howdy folks, sorry was late
16:03:02 sean-k-mooney o/
16:03:11 bauzas let's start
16:03:23 bauzas #topic Bugs (stuck/critical)
16:03:24 masahito o/
16:03:32 bauzas #info No Critical bug
16:03:39 bauzas #info Add yourself in the team bug roster if you want to help https://etherpad.opendev.org/p/nova-bug-triage-roster
16:04:09 bauzas any bugs to talk ?
16:04:56 bauzas looks not
16:05:02 bauzas #topic Gate status
16:05:12 bauzas #link https://bugs.launchpad.net/nova/+bugs?field.tag=gate-failure Nova gate bugs
16:05:17 bauzas #link https://etherpad.opendev.org/p/nova-ci-failures-minimal
16:05:27 bauzas (I'll leave the next PTL to curate that long list)
16:05:34 bauzas #link https://zuul.openstack.org/builds?project=openstack%2Fnova&project=openstack%2Fplacement&branch=stable%2F*&branch=master&pipeline=periodic-weekly&skip=0 Nova&Placement periodic jobs status
16:05:40 bauzas #info Please look at the gate failures and file a bug report with the gate-failure tag.
16:05:46 bauzas #info Please try to provide meaningful comment when you recheck
16:05:56 bauzas for me, the gate is correct
16:06:30 bauzas yet a few rechecks because of the usual suspects (guest panic and ssh timeouts) but nothing new and putting epoxy-3 at risk
16:06:39 sean-k-mooney bauzas: there is perhaps one

Earlier   Later