Istio Technical Oversight Committee, 19 Jul 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Technical Oversight Committee 2021/07/19

Description

Istio's Technical Oversight Committee for July 19th, 2021.
Topics:
- Review InsecureSkipVerify in DestinationRule
- Review WASM API

A

Remember to record this meeting, yeah yeah, all right, we're on we're live all right. um Shawty! Do you want to run through this or yeah.

B

Yes, so uh feature feeds is done. We have the first build. We started with the asynchronous testing. Last week, uh good news. We have two automation tests, thanks to doug bad news. We still have 20 tests which need automation, and this is the only week left for automation, testing, which means, if they are not done, the release will be blocked, as we normally do for p0s um and that's on automation, just to bring for the manual testing. There were a lot of manual tests.

B

I reviewed with the working group leads and most of them because they did not change from last release to this release. We marked them as p1. Only one of them was changed, which was related to security, so oliver and the team took it and they are doing the manual testing. So at this point honestly, all we have to do is automation testing and I really need help to automate those 20 tests which are remaining, and only one week is left.

A

So one thing when we first brought the subschweda there were some concerns that a lot of these were already marked somewhere as not applicable or something similar, um and that a lot of them weren't actually relevant, like I think.

B

I don't know so we open and review because we have filtered it on the ones which are not happy so go to the first yeah.

A

uh Yeah, like traffic management- this is just a.

A

This is just like a concept dock, so I don't think like this would ever be automated and I think it was on all the community testing sheets before we've marked it as not automatable.

C

Okay, we just need to mark it in istio.io as not automatable and it will fall off the list.

A

Yeah the same applies for almost all of these. I think, like this one oops, I.

C

Mean I don't know if some of these have snippets that need testing or not.

A

I yeah, I can say, like anecdotally, I know these three, all of these five or so six. They definitely don't. Maybe these other ones do we'd have to look closer, but I think a lot of them may actually not be required right. I mean.

C

A

The right spot to market eric, because I know that we've marked this various times in various places and we have.

C

A

C

Header of the document, uh oh of the actual doc, on easter dot, io correct! That's where brian and I generated this stuff from okay.

B

So marking them in the uh sheets before john, but this is literally from the steel.io website. So, as eric said, we have to change there and not in the sheet. So would you be okay to help us out there, because right now I have a list of 20 if we can reduce that will be awesome, because we have only one week left um and I really want to make sure we have the right test cases to be automated, which can be automated. That would be great.

A

Yep, that sounds good yeah. I think there'll be more than zero, but much less than 20.. Okay,.

B

So we have a working group leads meeting tomorrow um if you have time and just mark them in in the header um and here so we can review it tomorrow, based on how many are left, so I can distribute it accordingly.

A

B

So is it action item for you, john? If you can help.

A

Yeah all on that awesome. Thank you.

A

The other question I had about this is right. Now we took all the docs and then kind of split them up between different working groups.

A

um I'm just wondering, like it seems nice that everyone's helping out everyone else, that's good, um but it also seems bad that, like a working group which invested a lot of effort previously into getting these tests, now just has to keep doing more testing for working groups who focus maybe on features instead of testing um like it seems like if the working group hasn't tested all their docs, maybe their roadmap should be more focused on testing and they should invest over there instead of pulling in from other working groups, I'm not sure that there's going to be enough, docs that that's that that's a big deal once we filter out the ones that are not applicable, though, but I was just.

C

I thought I had well, I I was, I would say what we could do. John is um I mean we just took the top whatever number of docs that showed. You know that they didn't have tests and were the ones that were hit the most right. So we could take. You know 10 out of here 15, whatever the number is, but for the next release I mean we could try to be more proactive and saying you know. Looking at the last release.

C

This is what we think the top hit docs are and create this list and then get the work groups to do. What you're suggesting would that make sense.

A

Yeah, I think this list is super useful. It's just like if you look online 16 like this is a networking doc and it says user experience should automate it.

A

You know if the networking team can't document their stuff, then we they need to spend more time on testing and documentation instead of whatever other things they're working on right.

C

Right and I would pick on the same thing with tnr, because they got assigned some that weren't theirs either, but.

A

C

But yeah, maybe we could try to be more proactive either. You know, obviously we can't be this time around, but maybe we can be more proactive next time around and and see if we can identify the ones that we want automated earlier, especially as teams are working on road maps now, yeah.

B

Well, technically, all of them were automated right, um but then we changed the criteria and we just this quarter after the road map. We wanted to use the one which is heavily visited and this new uh test cases appear right. So what we can do, as johnny suggested, let's review this list right, which technically should reduce based on some of them or index and may not be automated, I'm hoping they will reduce from 20 to somewhere around, like 708, based on how you are going through that right.

B

That should be like at this point p zeroes- and this is the list where we said anything. More than thousand um visits um are our p zeros and to me after this quarter, all the p zeros should be tested. Right remaining will be p, one p two that should then we can distribute according to the road map and to which area they belong to proactively.

C

Right yeah pick the p ones or whatever you know the next 20 in the list, whatever right.

B

Yeah then we can start doing the p ones, but let's for p zero. I think that's the best approach we can do sounds good, yep, okay, okay, awesome. So the action item here right now for you, john, if you can help right as we discussed, adding to the header and the sheet, so we can review in the working replace tomorrow. That will be great.

B

Okay! Thank you. um So that's on testing. uh Now, looking at the bugs, there are two release blockers. um Of course we just came out of the weekend, so I do not have a good updates of if we have any mitigation plans or um what's the progress. So if you can go into that filter john on the release updates, yeah is john here I know brian is off any other release. Manager.

D

Okay, um are the release manager invited to this car.

B

I don't know, but I was assumed yeah.

A

B

Wouldn't think ryan is.

B

Okay, um then I'll work offline with them. There are two release: blockers and 19 p0s, which are scary at this point, um but I don't even know their progress to be honest if they are being booked upon or there is any. If anybody else know about the release- brokers at least speak up now or I can take it after.

A

um I know the second one: well, I created both of them. Actually, the second one is just mostly looking at envoy release, notes very hard and making sure that it's not going to break us. um So it's not necessary that something's broken it's, that we should make sure that something isn't broken because that's hit us a lot in the past. I took a quick glance and I'm nothing was terribly concerning, but I would love for a lot of people to have eyes on those um the other one.

A

We can get someone from from networking to look into it. I'll make sure that happens.

A

Okay, awesome, thank you. It should be small changed so.

A

B

A

P zeros or just those two release blockers.

B

um I wanted to, but since the release managers are not here, john um okay, I don't have updates so we can skip and we can discuss in the release channel or in the working repeats meeting tomorrow.

A

Okay sounds good.

B

A

Okay, um so 111 release managers. I think that's done. We can probably mark this off um 112 release managers. um It sounds like we have one from google. Do we already know who that is, or we just know we want one from google.

B

We do have one from google I'll, add the name confirming one more time, because I got for 111. I want to confirm they're, okay with one call too okay, but we need two more from other areas.

A

Yep sounds good. I think we we got the 111 release. Managers a bit late, so it'd be good to get the 112 ones started. You know as soon as possible. So if anyone else any other companies have volunteered feel free.

A

Okay, next, I think uh keynote added. This are you here.

E

Yes, um I am basically asking to have a review of the insecure skip, verify pri I put up and then ultimately, if possible uh it to be added as a feature into 111. But I understand if it's ability for that as well.

F

So can you have, I think, my approval on the api design. At least we require one more toc, member or any other member. At this point, it is all approved through the working groups right. So hopefully this is quick. What about the implementation side?

F

Do you have a pr out? I think you do. I just don't remember.

E

I have a portion for the environmental variable side of the original document, um but the actual insecure skip verify.

E

um I think I originally had something out, but I think it's just too out of date to to use it's more of a basically just checking and then removing any verification if uh insecure skip verify is true.

F

Okay, so you think for 111 like by when will you be able to collapse this, like that's the.

E

I'm assuming by this week, um but if I need to just do the api portion of insecure skip verify first, I could probably do that. I could definitely do it this week because I know the environment variable portion should also be backported into 110 and earlier as well.

F

Okay, so you should be able to so: let's try to get approval for the environment portion normal variable version of this, and then both the api and the implementation for insecure skip verify depending on when you get all of those things closed. The release managers will take a call.

E

F

But I think your basic request here is toc members, take a look at the api apr and if you can collapse, it.

E

Yeah, ultimately, that is um the biggest request, is just asking for a review.

E

Okay, I'll definitely take a look after the meeting.

A

E

F

E

A

I think at a high level it sounds fine to me, but I actually look at the pr.

E

Awesome. Thank you very much.

A

All right next looks like working group roadmaps um starting next week. Environments um telemetry still needs to schedule theirs. At some point um I don't know, I think doug. I think you mentioned you're out of office for a bit if that impacts it, but at some point feel free to fill us in yeah and.

E

Dart's gonna.

A

Take over for that, okay, cool.

A

Okay, is there anything else, your sweater, no.

B

It's just live for toc that we need to a lot some time each of this week for the presentations.

D

B

Yeah, thank you.

A

All right, um that's it! That's on the document. Is there anything else we need to discuss um yeah the the.

G

The wasm api pr it just needs, I think one more lgtm, uh but I mean like. I would really really ask people to do it in the next 30 minutes, because it's it's really done it's just like, should be done now.

C

I'll I'll take a look after this meeting.

D

C

G

D

On the priority.

G

Yeah yeah, so it's yes, so so they so we will update the dock with just explaining what it is right now, but it yeah it won't be. We won't change it. It's just. uh It is basically what you what you suggested just update the dock.

D

Okay is the priority higher means. It's earlier.

B

Yes, priority higher yeah.

D

A

And mandar is that in terms of implementation, is that uh ready to go with the pr and are we planning to add it into 1.11.

G

Yeah, so so the so the implementation uh is going to change a little bit based on some changes well to the api, but I don't expect the implementation to change a lot and the implementation doesn't destabilize anything else, because it's it's kind of opt-in by I mean you have to use the api and until the api is used, none of the code really kicks in, so uh that that's at least my my judgment. What's like.

G

What's your take on that john or or anyone else regarding the regarding accepting it in pilot at this stage,.

A

I think I'm generally fine with it because, like you said it's pretty isolated if it turns out that we need to make it, you know a bit more invasive and we start delaying this like a week or two. Then it may become questionable. But if you know we get the api merge today the implementation emerged this week and it's kind of isolated. That seems fine to me.

G

Okay yeah. I agree.

A

We should get make sure the release managers are aware and okay in the plan, though,.

G

All right, yeah, we'll uh we'll add, uh lag release managers on the on the master pr just so that they they're specifically expecting it.

A

Hey um anything else, we need to discuss.

D

So I do have a quick question. um Neuron on the api, you asked us to review the earlier one: insecure skip verify so the va. I saw there's a deprecation of verify. Client. What if I certificate a client, do we know of anybody using that it's the reason for deprecation is because it's moving to um config and proxy config, so you can configure it on a proxy basis.

F

uh It was never implemented and never documented.

D

I see, but it wasn't hiding from the dock. Okay,.

F

Yeah, as far as.

C

I know it was never.

F

Implemented only the api pr was merged. uh Canon do you want to jump in and answer.

E

Yeah, it was never implemented past the api being added.

D

There was an open pr.

E

That was never finalized.

D

Got you yeah thanks for clarification.

A

All right, um if there's nothing else, I guess we can end a review today.

F

All right see you folks next week. Thank you.

G

Thanks everyone.