[SIG-scale] Meeting Notes 2025-09-18
Andrew Burden
9/18/25
Recording: https://youtu.be/nbbpcvWv8m0
Attendees: rhallisey, fossedihelm
Agenda:
Review performance jobs
Node list calls increase from ~8 to ~35 around Jun 29, 2025
[alaypatel07] this seems to be coming from DRA feature
Create a unit test to verify the theory
This might not be a List Node call but a List ResourceSlice call
Sreeja, create an issue to see why kwok tests are failing to create primer VMI
Process did not finish before 4h0m0s timeout
Increase timeout
Make cluster-clean error
https://github.com/kubevirt/kubevirt/blob/main/hack/cluster-clean.sh
Script errors when it doesn’t find and is unable to clean up a VMI
[fossedihelm] This PR https://github.com/kubevirt/kubevirt/pull/15599 should increase some performance.
- Previously, for addition/updating the paused condition, we were patching the entire conditions field (the probability of a conflict is high, and hence re-enque of the key).
- Now, there should not be any conflict -> low patch count / lower workqueue size
- I remember seeing many times the `error patching pod conditions` log in the virt-controller logs
- Can we track this? And see if this is true?Reduced Patch Pod calls by 50%
Can we use strategic merge patch elsewhere?
[fossedihelm] Consider updating the provider to 1.32 or 1.34 in release-1.7 lifecycle
Need to upgrade provider from 1.31
Move to 1.32 and gather ~2 months of date to review
Make the change a few weeks after dedicated CPU change
~End of October