Kubernetes SIG Instrumentation, 20 Aug 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: SIG Instrumentation 20200820

Description

SIG Instrumentation August 20th 2020

A

Oh yeah, I can share the meeting today. uh Welcome everybody to the uh august 20th 2020 sig instrumentation meeting, uh I'm alana hashman. I am one of your sick chairs and I'm hosting today uh welcome everyone. I think this is gonna, be a pretty short meeting, because uh this week is kubecon eu, but I have not been attending that as it is not on my time zone. So I am here and ready to go uh and han do you have anything for today.

A

Otherwise I will jump into the agenda which I will link in the chat.

B

Just the one thing we did meet up.

C

I just want to let people know that we met up last week about triaging cigarette instrumentation issues, uh we're going to start a weekly triage, and so there should be another. There should be another calendar invite starting next week for a weekly triage on wednesdays. So if you are interested, uh please feel free to come in sign up for tickets uh yeah. That's it.

A

Great, I have put that in the meeting minutes. um Okay, I have one topic for today, uh which is uh end end test roundup, uh as I was going through various things uh assigned to us uh and or uh pinging the sig instrumentation group on github. uh I found that there were a bunch of random end-to-end tests that were failing. That uh folks were asking for attention on um a lot of these.

A

End-To-End tests seem relatively old and unmaintained, and I don't know if there's anyone in the sig who's currently like paying any attention to these things.

B

A

D

Elastic search.

A

Yeah, like the elastic search tests, so um there are three issues that I linked uh that we had been pinged on, uh like they were asked to be targeted for, like 119 at least one of them. I think, at this rate, that's not going to happen, uh but uh yeah. We have some failing elastic search tests under the uh the sig instrumentation end-to-end tests, uh there's one about some missing logging agents and there's one about.

A

um Sometimes we got a flaky test where, like a pod, is not registered, and so the metrics gathering get skipped, but that is an error and therefore like the test, fails because it expects no errors. So uh I looked into a few of these, like I guess the reason that I brought it to the sig as opposed to just like trying to crack away at these bugs is. This seems to be a theme like that we have a whole chunk of end-to-end tests that seem to be unmaintained.

A

So what do we want to do about those? uh If anybody has any ideas.

C

uh I would like to see let me.

D

D

D

D

uh This is many many years old.

A

I guess I'll ask: uh is there anybody on the call who's interested in uh stepping up to help maintain some end-to-end tests versus instrumentation.

C

Merrick, are you on the call.

A

Looks like yes.

C

Do you know anything about these? I know you know quite a bit about mugs.

E

So uh so there at the time that we maintain cuba that the original way to to to deploy kubernetes in the company district, kk repository.

E

We, we developed some add-ons and there were like two stacks one for stackdriver logging, second for elastic search and because no one maintains stack driver. I removed all of those. But there were some people that wanted to maintain elasticsearch like we.

E

I didn't find it like enough to cause to delete them and ask them like twice to start looking into those tests, but they.

E

Shape, I'm also somewhat worried about like that, the security state of this add-on, because it depends like it's addon in kubernetes, kubernetes repository that is not really maintained, uses image based images with security vulnerabilities, and we cannot do and we don't do anything with it.

C

ah That sounds like an issue.

A

Yeah uh there was a bit of a discussion in the slack channel a bit back about the elastic search add-on. I think in particular, uh because there's some concerns that the tests have just been like failing forever and it seems to be quite unmaintained like I could not see any active development on it, like uh other than you know, like small maintenance stuff since, like 2017, or something like that, um so I am all in favor of removing unused code, particularly if it is like not working like if the tests are all just failing.

A

They have no value, um so I would be in favor of getting rid of that stuff. At least although there are other tests in the end to end test suite. I think that are not just the elastic search ones, but.

C

It's testing something right like it's testing some bit of code like if you can't just delete the test and not remove the code.

A

No, no I'm saying we should also delete the code uh for the yeah. Actually, okay, so two of the three uh are both under the uh the logging tests uh for, like the end to end stuff, so uh one is for elastic search. The other is for stackdriver.

A

uh I made a little bit of a face when I saw the stackdriver one, because I mean this test is so old that, like it's using the old terminology for worker nodes like it's calling them minions, this stuff has not been maintained so, um like I think, it's probably safe to just rip those tests out but like what we want to generally do about the elastics like we should also go and like possibly rip out any of the code. That's referencing these tests.

A

C

Mean okay, so the stackdriver stuff I feel, like probably merrick, has the best context on that. That's my guess. Given.

E

All right I can make decision like I can find someone who can make decisions that we can delete or someone that would mainta maintain like fix them so for skydriver for all of them. First driver no, like, like.

E

I know the person that out of the stack driver- uh and I think like if it's okay, I could we could just fix them, but for elasticsearch like I've, never like outside of slack, I don't have really good contact with them and they never.

E

I was never able to convince them to to fix anything.

C

Okay, then, then, uh we'll just create an action item for you to follow up on the stack driver. uh I guess we have to figure out what to do about the elasticsearch thing, but it doesn't look like people have touched this in a really.

D

C

I don't even know what this test elastic search logged in.

A

So merrick I put an action in the minutes uh for you to find someone to help out with the stack driver tests. uh Can I assign that github issue to you.

E

B

E

You cannot find this type of.

A

Okay and I'm I'm not sure yet what we're gonna do with the uh elasticsearch.

F

C

uh Maybe we should okay.

C

D

Comes to my issue.

C

If someone creates an issue in a science thing, I can try to find somebody.

D

Who touches this? What.

A

D

A

I think so the elastic search uh and end test that one is currently assigned to me. uh We cover that one, I think in our first triage session um or uh like it was otherwise mentioned uh on site.

E

A

For the uh the metrics grabber, one which is not doesn't have anything to do with the logging stack, I think it's just a bug in the test. I'm not really sure who I can try to maybe chase down someone yeah.

C

That's controller manager uh that should yeah, I I know I I know who to assign that to that's. Not us. That's controller manager.

A

Okay, uh then do you want me to assign that one to you, so you can find someone to work on that. One.

G

Sure, okay, it I looked at that one really briefly. I think it was one flake in like three months or something so it's. It was just one instance of the test where the controller manager pod didn't come up.

C

G

So there may not be that probably close.

C

It we can probably close it.

A

Oh okay, uh let it I'll I'll add a comment then, to that one saying that you did that research and uh we don't think it is. uh That was just one flake.

G

Well, yeah: it's a flaky test, but I don't know if there's that much there.

D

D

A

Okay, uh that's easy, then uh cool. So then I guess we've got um I'll close that one and then we've got uh america signed to follow up on the stackdriver one and then um han I'll assign you the uh the elasticsearch one, okay sure, uh and then I think at least that covers I mean we haven't gone and done like a triage for any issues. Yet uh as leads, uh our first meeting was more of a meta meeting.

A

uh We were mostly focusing on prs, uh but that's something we can do in the future uh and try to like chase down these various things, uh but that at least sort of plugs the hole for uh those things. So that's all I had on that topic uh unless anyone else has anything to add.

D

C

How is it how's the keep going going.

E

And I only attended my session the rest of the. I will watch them when they upload the videos when they're uploaded.

C

They're not uploading cool, okay,.

B

You should you should share them when they're uploaded.

C

Yeah sure, all right, um I I don't really have anything, there's anyone else or.

C

Should we call it 15 minutes early.

A

I'm happy to call it, but it looks like sergey has something yeah.

H

I just want to introduce myself, I'm sergey, I'm uh I'm new in google and I'm working on signal and uh history like in my past is in open telemetry. So I'm in governance and technical committee and open telemetry and I finally uh figured out my daycare situation. So um I can attend this meetings. um I will try to be flying a wall for a couple meetings, but I really love all these things cinematography. So I hope I will contribute a little bit here.

A

Awesome welcome survey.

H

Yeah, thank you.

A

Okay, we have anybody else on the call, who's new and would like to say hi.

F

Hi, I I think, I'm going to say hi as well, I'm from usa- and I I'm sorry so I just put in my ipad walking around so I started working on the observability uh for answers since uh january this year. I just attend this meeting and try to learn more things, and I I was about to say that the the stackdriver related test, I am willing to help so yeah. That's all awesome.

A

Awesome great nice to meet.

D

A

David, do you have anything.

G

um No I'll, I I have a thing I'm gonna share, but maybe I'll I'll wait till a different week.

A

D

C

uh Cool then uh have a nice two weeks. I will see everyone in two weeks or next week. If you come to bug scrub, which is going to be on wednesday.

A

Great I'm going to stop the recording thanks for attending everyone.

D