Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2026-07-20
16:09:00 fwiesel Yes, I looked into the pipeline failures, and instance creations were slowed down to that extend that no instances were being created.
16:09:32 fwiesel #link https://review.opendev.org/c/openstack/oslo.vmware/+/997269
16:10:01 fwiesel This should reduce the overhead of doing api calls against VSphere, but I am suspecting it is not the full picture.
16:10:09 fwiesel Still digging into it.
16:10:13 fwiesel That's from my side.
16:11:24 bauzas cool thanks
16:11:31 bauzas #topic Kamil's news about eventlet removal
16:11:41 bauzas sambork ?
16:11:52 sambork o/
16:12:20 sambork last week was dominated by test stabilization: the unit test exclude list is gone, and there's a push to make the threading functional CI job voting.
16:12:53 gibi e.g. https://bugs.launchpad.net/nova/+bug/2161293 ^^
16:13:07 sambork he code cleanup is moving forward by replacing Nova-internal threading utilities with futurist primitives (both nova and futurist patch are pushed and ready for review)
16:13:31 sambork eom
16:13:51 sambork anything to add gibi? (thanks for bug above)
16:14:14 gibi fwiesel's finding is sort of eventlet related and I made mental note about performanc of the oslo retry decorator that could effect nova as well not just oslo.vmware
16:14:38 gibi we might want to take a hard look of the usage of that and consider tenacity's cheaper retry all over the place
16:15:32 gibi aaand I still haven't filed the bug to switch the default of our unit test execution to threading and just keep one tox job for unit test + eventlet
16:15:47 bauzas that's nice we no longer have unit test exclude list
16:15:47 gibi but I will I promise... eventually...
16:16:20 gibi btw the functional threading job has only 2 test case excluded
16:16:26 bauzas nice
16:16:42 gibi and at least I understand the problem in one of them :):
16:16:45 bauzas nice
16:17:30 sambork do we need additional ticket for checking retries? and do we wan to priorities this?
16:17:52 gibi sambork: please file a bug just to not forget it
16:17:59 gibi right now I don't feel it has a high prio
16:18:09 sambork ack
16:18:17 gibi but it might become one based on performance results from downstream testing
16:19:00 opendevreview Merged openstack/nova master: Use domain capabilities to detect SEV-SNP support https://review.opendev.org/c/openstack/nova/+/996316
16:19:13 opendevreview Merged openstack/nova master: Extend tests for _is_supported_mem_encryption_model helper https://review.opendev.org/c/openstack/nova/+/997663
16:19:26 opendevreview Merged openstack/nova master: Refactor CPU traits functional tests for AMD SEV support https://review.opendev.org/c/openstack/nova/+/997635
16:19:34 gibi that is basically the report for the SNP status ^^^ :)
16:19:49 antia perfect timing :)
16:19:52 gibi yeah
16:19:59 gibi we can move on
16:20:39 bauzas ack
16:20:45 bauzas #topic Nova using openstack sdk for neutron
16:20:49 bauzas who is leading that ?
16:21:20 elodilles this is Lajos' topic if i'm not mistaken and he is still on PTO
16:21:32 bauzas cool moving on then
16:21:54 bauzas #topic Ironic review priorities
16:21:59 bauzas anything to say ?
16:23:43 bauzas looks not
16:23:48 bauzas #topic Confidential computing, status of SNP and TDX features
16:23:53 bauzas I've seen good news
16:24:27 tkajinam the remaining sev-snp patches lust landed so sev-snp support was completed
16:24:33 tkajinam \o/
16:24:40 gibi \o/
16:24:44 tkajinam I just noticed that https://blueprints.launchpad.net/nova/+spec/amd-sev-snp-libvirt-support is not yet approved. can someone do so ?
16:24:54 tkajinam then I'd update its implementation status
16:25:24 gibi done
16:25:27 antia I suppose the same for https://blueprints.launchpad.net/nova/+spec/intel-tdx-libvirt-support
16:26:31 antia anyway I ended up with PTO last week so not much on TDX side
16:26:31 tkajinam because these changes were merged tdx needs another rebase but I think we are now fully ready to start it.
16:26:38 gibi antia: fixed the bp for that too
16:26:48 antia gibi: thanks!
16:27:08 gibi I will grab a TDX hardware this week (hopefully)
16:27:22 antia started with some catch up with SNP and rebased with master, noticed some changes to the test structure which results in errors for me
16:27:29 antia I suspect it’s because the exists mock that was dropped in https://review.opendev.org/c/openstack/nova/+/994343/18/nova/tests/fixtures/libvirt.py
16:27:36 antia Since this is the error "Exception: This test invokes builtins.open on /sys/module/kvm_intel/parameters/tdx. It is bad, you should mock it."
16:28:00 antia gibi: Cool! Will get you something to test :)
16:29:17 antia In the end the sys fs check is not needed for TDX since Libvirt does it anyway so the easy fix is probably to drop it, but mostly curious if it was solved in some other way for SEV
16:29:36 gibi antia, tkajinam re mock, yeah we should either mock open or have a different way to check the status
16:29:52 gibi tkajinam: did we fully removed the reliance on the kernel module parameter?
16:30:17 gibi if so, and that worked for SNP, I hope we can do the same for TDX as well
16:30:33 tkajinam gibi, no we have to rely on it for snp but we can drop it for tdx
16:30:49 gibi tkajinam: due to the ES/SNP CVE?
16:31:09 tkajinam no. due to libvirt not checking the flags for es and snp
16:31:19 tkajinam we could probably get rid of access to sev flag, though
16:31:42 antia I'll just drop it as part of the rebase and we should be good
16:31:46 gibi tkajinam: antia: so for TDX we could go with the domain capabilities route?
16:31:51 tkajinam I mean it only checks /sys/module/kvm_amd/parameters/sev and does not check /sys/module/kvm_amd/parameters/sev_es or /sys/module/kvm_amd/parameters/sev_snp
16:32:19 tkajinam gibi, yes, libvirt checks the tdx flag when it declares <tdx support='yes'/>
16:32:33 antia gibi: Yes exactly. Right now it does both the sys fs check and domain capabilities which is not needed
16:32:51 gibi cool then you can get rid of the problem by not relying on reading that file :)
16:33:15 tkajinam yeah using only domain capabilities simplifies the logic and tests largely
16:33:51 antia tried it before and tests were quite annoying to fix, but I think that has been fixed already so should be simple enough
16:34:20 gibi cool
16:35:03 tkajinam one question I got during reviewing the tdx series is whether we need to bump object version again for tdx or can reuse the existing version for snp assuming tdx would be merged this cycle
16:35:21 tkajinam I mean for snp we bumped ImageMeta object version due to hw_mem_encryption_model supporting new amd-sev-snp
16:35:46 tkajinam and I'm wondering if we should bump the version again when hw_mem_encryption_model supports new intel-tdx or we can handle it within the same version
16:36:45 gibi bump the version please
16:36:50 gibi object versions are cheap
16:36:56 tkajinam ok
16:37:01 antia either is fine with me
16:37:05 antia will bump it
16:37:07 gibi thanks
16:37:30 tkajinam quick conclusion :-)
16:37:44 gibi :)
16:37:50 tkajinam nothing else from my end. I reviewed tdx sceries and left a few comments there.
16:38:03 antia thanks for the review!
16:38:04 tkajinam antia, in case you need help to follow existing sev thing then feel free to ping me
16:38:16 antia will do :)
16:38:28 gibi I will keep myself in the loop :)
16:39:59 gibi we can move on
16:40:55 tkajinam yup
16:42:05 gibi bauzas: are you still with us?
16:43:10 gibi OK I'm taking over :)
16:43:15 bauzas sorry back
16:43:22 bauzas #topic Bug scrubbing

Earlier   Later