Friday, 2024-02-23

*** haleyb is now known as haleyb|out00:58
opendevreviewmelanie witt proposed openstack/nova master: Report ephemeral disk encryption in the metadata API  https://review.opendev.org/c/openstack/nova/+/90994501:50
opendevreviewmelanie witt proposed openstack/nova master: Deprecate legacy ephemeral storage encryption using dm-crypt  https://review.opendev.org/c/openstack/nova/+/90994702:26
melwittlooks like nova-ceph-multistore is fixed by the merge of https://review.opendev.org/c/openstack/requirements/+/90992302:54
opendevreviewMerged openstack/nova master: [S-RBAC] adapt nova-next for port's binding:profile field change  https://review.opendev.org/c/openstack/nova/+/90985905:31
opendevreviewMaxim Monin proposed openstack/nova master: Server Rescue leads to Server ERROR state if original image is deleted  https://review.opendev.org/c/openstack/nova/+/87238505:32
opendevreviewmelanie witt proposed openstack/nova master: Deprecate legacy ephemeral storage encryption using dm-crypt  https://review.opendev.org/c/openstack/nova/+/90994705:52
opendevreviewMaxim Monin proposed openstack/nova master: Server Rescue leads to Server ERROR state if original image is deleted  https://review.opendev.org/c/openstack/nova/+/87238507:42
opendevreviewDr. Jens Harbott proposed openstack/nova master: DNM: Test glance new location api  https://review.opendev.org/c/openstack/nova/+/89120708:03
*** mklejn_ is now known as mklejn08:24
LarsErikPI have discovered something I wonder if it's a bug, or if it is intended behavior. In yoga at least, I can unshelve a VM that will cause an overconsumption of the CPU quota..09:36
LarsErikPCreated a project with 4 VCPUs and 4GB RAM. Then spawned server1 with 2 cores and 2GB RAM. Shelved it. After that i spawned server2 with 4 cores and 2GB RAM. I have now used 4 of 4 CPUs in the quota. Then I unshelved server1 without any problem, and now I have used 6 of 4 CPUs...09:38
LarsErikPI.. did not think this should be possible, as it enables users to easily overconsume their quota..09:39
LarsErikPyou can do the same trick with overconsuming RAM as well09:48
LarsErikPand, fwiw, both at once is also possible09:51
bauzasLarsErikP: most of the community today is on PTO10:33
bauzasare you sure that your limits were for 4 VCPUs ?10:34
LarsErikPyes10:48
LarsErikPI filed a bug anyways :P https://bugs.launchpad.net/nova/+bug/205479710:49
LarsErikPand it kind of scales.. meaning if I create a bunch of VMs, shelve all of them, create a new one consuming the entire quota; I can stil unshelve all the shelved VMs10:52
opendevreviewsean mooney proposed openstack/nova master: add pyproject.toml to support pip 23.1  https://review.opendev.org/c/openstack/nova/+/89975311:06
opendevreviewKonrad Gube proposed openstack/nova master: Use Cinder's os-extend_volume_completion volume action.  https://review.opendev.org/c/openstack/nova/+/87356011:20
opendevreviewsean mooney proposed openstack/nova master: Revert "[S-RBAC] adapt nova-next for port's binding:profile field change"  https://review.opendev.org/c/openstack/nova/+/90983411:55
kgubebauzas, gibi: Hi, could you have another look at this? https://review.opendev.org/c/openstack/nova/+/87356012:47
opendevreviewMerged openstack/nova master: Follow up patch to correct a minor documentation issue  https://review.opendev.org/c/openstack/nova/+/90886313:26
opendevreviewMerged openstack/os-vif master: Fix missing or unnecessary dependencies  https://review.opendev.org/c/openstack/os-vif/+/90934114:21
u00599Hey, we came across a problem in our openstack environment where the metadata-service is not marking interfaces with dhcp_enabled for network config with cloud-init. We came across the following commit which wanted to solve this problem, but this doesn't work for ovn with native dhcp (no agents) as there are no dhcp-agents anymore. The concerned networks don't have a default gateway set, so the "fix" is not working for us. Anybody came 14:49
u00599https://github.com/openstack/nova/commit/c7f572d65b57e034c1391ed49d84f6e5f1d672ad14:49
artomSo it's hw:emulator_threads_policy with the s but hw:cpu_policy without the s15:17
artomWe're just evil, aren't we15:17
opendevreviewMerged openstack/nova master: check both source and dest compute libvirt versions for mdev lv  https://review.opendev.org/c/openstack/nova/+/90417615:40
jamespageLarsErikP: https://bugs.launchpad.net/cloud-archive/+bug/2054687 to track15:40
opendevreviewMerged openstack/nova master: Check if destination can support the src mdev types  https://review.opendev.org/c/openstack/nova/+/90417715:46
opendevreviewMerged openstack/nova master: Reserve mdevs to return to the source  https://review.opendev.org/c/openstack/nova/+/90420915:46
sean-k-mooneyu00599: ovn has its ovn metadata agent15:52
sean-k-mooneyu00599: but unless they have fixed it neutron is not waiting for it to be ready before seneding the network-vif-plugged event15:53
sean-k-mooneyso there is or was  a race when using ovn15:54
sean-k-mooneyLarsErikP: this is related to counting limits in placment16:00
sean-k-mooneyhttps://docs.openstack.org/nova/latest/configuration/config.html#quota.count_usage_from_placement16:06
sean-k-mooneyBehavior will be different for servers in SHELVED_OFFLOADED state. A server in SHELVED_OFFLOADED state will not have placement allocations, so it will not consume quota usage for cores and ram. Note that because of this, it will be possible for a request to unshelve a server to be rejected if the user does not have enough quota available to support the cores and ram needed by16:06
sean-k-mooneythe server to be unshelved.16:06
sean-k-mooneyLarsErikP: https://review.opendev.org/c/openstack/nova/+/888096 is the yoga fix but yoga is now unmaintained https://opendev.org/openstack/nova/src/branch/unmaintained/yoga16:16
sean-k-mooneyhum although im not sure that is corect since wallay and victoria should be unmainted then too16:17
sean-k-mooneyand xena16:17
sean-k-mooneyelodilles: ^ we should look into what up there16:17
sean-k-mooneyhttps://opendev.org/openstack/nova/branches16:18
sean-k-mooneyi forget what we had agreed but i taught we agreed not to skip any releases16:18
sean-k-mooneyso if yoga is moved to unamined i woudl expect all older branch to be as well16:18
elodillessean-k-mooney: we started the transition with yoga, as that had a 'deadline' some months ago... now we'll continue with the rest (xena, wallaby and victoria), actually, i just said on the Release meeting that i'm planning to propose the transition patches on Monday (despite the busy Caracal-3 milestone busy time... :S)16:29
sean-k-mooneyelodilles: oh ok16:30
sean-k-mooneyso they are just in flight16:30
sean-k-mooneycool16:30
elodillesyepp-yepp16:31
elodillesthat's the plan16:31
artomBleargh, the CPU power management stuff is not testable for multiple compute hosts in functional tests16:52
artomAll the work is being done by module-level functions, no way to mock/patch different instances on the two compute services/libvirt drivers16:52
artomIOW, live migration reproducer tests are impossible16:52
artomAnd I feel like if I make the CPU API a class and stick in an attribute of the driver, it won't fly, because it's a change that would only exist to allow tests16:53
sean-k-mooneyartom: we could test it with multiple compute nodes in the functional tests16:54
sean-k-mooneybut we are missing test infra to do that16:54
artomsean-k-mooney, I don't see a way to the assert the online/offline CPUs on each host16:54
artomBecause the libvirt driver imports cpu/api.py as a global libvirt_cpu, and calls that16:55
artomActually, what I'd really need is to somehow stub in different "instances" of the Core class for each libvirt driver17:03
artomMore like, a different SysFS fixture for each...17:04
sean-k-mooneyartom: we made live migration work in teh functrional tests with sriov17:17
sean-k-mooneywe can make it work for power management17:18
sean-k-mooneyif we put in the effort17:18
artomRight, I'm just not sure it's possible without changing the actual Nova code17:18
sean-k-mooneyanywya im ment to be off today so im going to leave the office again17:21
artomYeah, I was going to say, make sure you compensate for working today by not working Monday or something :)17:22
elodillessean-k-mooney LarsErikP : i've cherry picked the patch to unmaintained/yoga: https://review.opendev.org/c/openstack/nova/+/91001017:47
elodillesgate is happy with it, though there are no devstack based jobs among the zuul jobs at all, so that should be added to the branch first17:47
opendevreviewArtom Lifshitz proposed openstack/nova master: Power on cores for isolated emulator threads  https://review.opendev.org/c/openstack/nova/+/90979518:14
opendevreviewArtom Lifshitz proposed openstack/nova master: POC: power up dedicated cores during pre_live_migration  https://review.opendev.org/c/openstack/nova/+/90980618:14
opendevreviewArtom Lifshitz proposed openstack/nova master: Add cpuset_reserved helper to instance NUMA topology  https://review.opendev.org/c/openstack/nova/+/91002018:14
opendevreviewArtom Lifshitz proposed openstack/nova master: Reproducer for not powering on isolated emulator threads cores  https://review.opendev.org/c/openstack/nova/+/91002118:14
opendevreviewArtom Lifshitz proposed openstack/nova master: WIP: Reproducer tests for live migration with power management  https://review.opendev.org/c/openstack/nova/+/91002218:14
u00599sean-k-mooney: but infact ovn has just it's proxy and network_data.json is provided by nova-api or am i wrong? Infact the ovn has it's router and dhcp port in common and is owned by network:distributed. Wouldn't it be feasible to add another check for this owner when no dhcp-agent is found?19:48
opendevreviewJay Faulkner proposed openstack/nova master: Limit nodes by ironic shard key  https://review.opendev.org/c/openstack/nova/+/90391522:55
opendevreviewJay Faulkner proposed openstack/nova master: Add nova-manage ironic-compute-node-move  https://review.opendev.org/c/openstack/nova/+/90391622:55
opendevreviewJay Faulkner proposed openstack/nova master: Make compute node rebalance safter  https://review.opendev.org/c/openstack/nova/+/90391722:55
opendevreviewJay Faulkner proposed openstack/nova master: Make compute node rebalance safer  https://review.opendev.org/c/openstack/nova/+/90391723:23

Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!