[jed] couldn’t find the source of the issue in the pprof files, but maybe someone else with actual pprof skills could find something
[jed] this issue seems quite critical, we need more eyes on it
Please keep sig-scale in the loop with cause of this for their tests
Worth adding an alert for gross mem usage?

Flaky test fixes: https://github.com/pulls?q=label%3Akind%2Fflake+repo%3Akubevirt%2Fkubevirt

[Tests]Use Eventually in test to allow time for reconciliation loop: https://github.com/kubevirt/kubevirt/pull/13332
vsock tests: use pgrep instead of non existing lsof: https://github.com/kubevirt/kubevirt/pull/13330
fix(tests,vsock): Attach /dev/null to netcat's stdin: https://github.com/kubevirt/kubevirt/pull/13409

Zoom chat:
00:04:37 Andrew: https://docs.google.com/document/d/1nE09vQWcCTW-9Ohe9oCldWrE0he-T_YFJ5D1xNzMtg4/edit?tab=t.0

00:13:20 xpivarc: Thanks Alay!

00:14:46 Alay Patel US: Reacted to "Thanks Alay!" with 🎉

00:15:48 xpivarc: We want to migrate one proposal in January and one in February. So they would target 1.6

00:16:37 afrosi: I have read the proposal but haven't commented yet. Why no finalizer?

00:18:26 akalenyu: Maybe this problem is general to more resources in a cluster so possibly some higher level operator would protect against it?

00:18:45 Adi Aloni: Reacted to Maybe this problem i... with "➕"

00:25:50 xpivarc: Have we fix opened for the little/big edian issue?

00:26:36 vladik: Thanks Alay!

00:28:09 akalenyu: Replying to "Have we fix opened f..."

I proposed a clear path forward using the C based "blockdev" util for the short term but I don't know if the s390x folks will give it a go or not

00:29:06 xpivarc: Replying to "Have we fix opened..."

Do the raw syscalls equivalents suffer frowm the same issue?

00:29:15 akalenyu: Replying to "Have we fix opened f..."

00:29:24 akalenyu: Replying to "Have we fix opened f..."

This is why blockdev works

00:29:41 xpivarc: Replying to "Have we fix opened..."

So let's just use them or not?

00:30:02 akalenyu: Replying to "Have we fix opened f..."

Possible

00:30:06 vladik: not me ..

00:30:34 akalenyu: Replying to "Have we fix opened f..."

I would prefer the extra abstraction via blockdev so kubevirt doesn't have to appear to know so much

00:37:41 Itamar Holder (iholder101): +1 for Reproduction steps

00:40:28 vladik: Alay, do you monitor the virt-handler mem consumption internally? (if it's not a secret ;) )

00:40:50 ffossemo: my headphones...

00:41:26 xpivarc: Replying to "Have we fix opened..."

C is harder to debug ;) just saying

00:42:28 Alay Patel US: Replying to "Alay, do you monitor..."

We do care about virt-handler mem consumption, dont recall top of the my head what kind of monitoring/alerting we have, let me get back to you on that

Reply all

Reply to author

Forward

0 new messages