Kubernetes SIG Node, 8 Jun 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Kubernetes SIG Node CI 20220608

Description

SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk

GMT20220608-170604_Recording_2084x1974

A

It is the 8th of June 2022, and this is the signaled CI subgroup weekly limiting. uh It looks like we have a pretty empty agenda for today, unless anyone has anything last minute.

A

No then cool. Let's move on to triage uh when it comes to test grid. It looks like things are mostly the same.

B

A

A

um We have a new failure for cryo.

A

Looks like the cryo tests have been flaky fall at least a few days.

A

Oh, isn't there already an issue for this? One create a sequence of PODS. Latency resource should be within limit when create 10 pods with 50 background pods I feel like I've, seen this test before.

B

A

We had issues in the container view version of this before.

A

um Does anybody want to take a look at this.

C

Yeah I could grab it cool.

A

um Performance wise looks like nothing is really changing. Just the usual like fun variants,.

A

uh I wonder what happened there, though? Okay one from using no CPU to using some CPU. That's fine.

A

And the test triage board is pretty lightweight this week. I just added the stuff before this call.

A

Skip a runtime class job for none GCE providers.

A

A

Reasonable to me.

A

A

A

uh Does anyone want to give this a quick sanity check to see if there's a better way we should skip or if we do it elsewhere,.

A

um The skipping, a test that requires um attained to exist um on anything that isn't GCE, which just seems a bit broad to me.

D

I'll look at it.

A

All right, yeah.

D

It's uh R Phillips.

A

On yeah yeah I cannot type today.

A

uh Couplet monk ready conditioners false explicitly for terminal parts.

D

This is in review at the moment, and uh Morel is looking at it already. Oh.

A

Yeah, it looks like it's also already triaged, but for the CI board that can be.

A

A

Oh, it's already been all gtmed cool.

A

And windows: I: don't think we really.

A

Look at too closely, but I will take a look at this one or get someone from Windows to take a look.

A

uh That's the test triage done so if.

B

A

Are only interested in tests and not bugs uh feel free to drop.

A

um But it looks like the bug list is pretty light this week, too,.

A

uh List pod error: after a pods evicted after I evicted, a part I performed the operation of the full list instance, but found something that did not meet expectations.

A

Cannot assign or convert an eviction into a poet, oh they're, using.

A

That seems weird.

A

B

A

A

Like not a notebook.

D

Yeah, it seems more like a API or.

D

The client issue.

B

E

At the top it says maybe fixed by and then I think that's a PR number.

A

I'm gonna link these and then uh API Machinery, I, guess.

D

Yeah, it looks like client go.

A

um Also, today, I learned about that feature in GitHub. That seems useful.

C

Yeah I've never seen that before that's kind of cool.

A

Github has been shipping like a bunch of awesome stuff lately.

A

uh Superfluous right ahead, our call.

A

A

um Does anyone want to take a look at? This uh looks like some.

A

Weird, like header, writing from lugs.

C

I could look into it and I'm, not an approver or anything but seems like a fun bug. Cool.

A

C

A

Your account handle yeah.

E

T stapler s-t-a-p-l-e-r yep.

B

A

I lose the ability to use a keyboard when I'm sharing my screen.

A

Stressful, uh no disk pressure is missing when Dev MacBook usage is full.

A

A

Calculating the image FS, but it's not a set disk pressure, nor effect pods.

B

A

I assume this one is just slow accounting.

A

A

This needs information because it doesn't show how the kublet is configured with.

A

Limits like do. We have good defaults for that.

A

A

A

um Looks like sweetie took a look at this and.

A

A

A

This is already assigned, but uh Francesco. It does not see through Nicole.

A

A

This looks more like.

A

A

And numerous Affinity configuration.

A

A

Looks like this is support, not bug.

A

And I think that, aside from.

A

Where is that yeah.

A

A

A

The conversation is going in tangents and not really answering what we need. uh Sandbox plus containers remain undeleted. Also another Docker bug.

A

uh Liveness per failing when the load on the application is maxed, they didn't get back to us, and this is also from last time. So that means everyone should get 40 minutes back.

E

Awesome see you.

A

Have a good rest of the week.

A