Antrea Community, 14 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Antrea Community Meeting 03/14/2023

Description

Antrea Community Meeting, March 14th 2023

A

So good morning or good afternoon, thanks for joining this meeting, um so this is today is Wednesday March the 15th, and we should be very close to the entria 1.11 release date before we start with today's meeting. I just have a quick, Service Announcement. uh This is the last meeting uh where I mean from next week from the net from the next meeting.

A

Also Europe will be switching to Daylight Saving Time, which means that, uh since our meeting is based on GMT time zone, the meet the meeting time will go back to normal for Forks in United States, but it will also shift by an hour for all those countries that do not adopt the daylight saving time so uh in an actual uh for India to shift from 10 30 to 9 30 a.m and uh from China from 1 pm to 12 P.M.

A

So uh please bear in mind that uh the next distance of the meeting May by be at a different time so in the meanwhile santolan join the good good evening. Thanks for joining, we I appreciate it's kind of late, and also you might be one of the lucky ones which is not affected by the power outage in the Bay Area.

B

Yeah this time, I'm lucky because two weeks ago, I lost power for three days. So.

A

Yeah, that's right! It's some! It's it's! Somebody else is starting.

B

Also lost power for two days last time and he doesn't have power now so.

A

Maybe janjun is the one that goes and cuts the cables Maybe anyway. So uh I would say that for today we don't have any uh pre-scheduled. We booked a meeting in the agenda, so uh we'll probably just won't like to go through an update for the status of the entry at 1.11 release, and then maybe we do uh open discussion to see if there is anything that the team will likely bring up so I don't know. Antoine Channel wants to go ahead with the an overview of the release status.

C

uh I could get a quick report about the current status of release 1.11 and currently we we have planned to delay the release three days so is in fact, if it's not expected to be released on next Wednesday I. Think and perhaps let me share my screen. We have to go over the current status.

C

Yeah I can't read that as the 16 PRS opened for this release, some of them are critical, bug fix and some of them uh new features introduced in this release.

C

Let's go through the list for the first one. I think this is a bad fix.

C

And the needed if we wanted to graduate the future endpoint slides from alpha to Beta, which means that after this graduation, the endpoint slides will be used by default by Andrew proxy and according to the test, there's one signal: walk test started to fail after this switch, so we have to fix the dark fast before we merge this PR and uh our second one.

C

It's not uh new requirement but kind of refactoring about how we handle uh the what the user input provided configuration and how we consume, then in the uh entry of the program, and because it may introduce a big scope of change, changes and would might be conflict with existing PRS. I I didn't update it in time and I suppose. If this is not, cannot be merged in this release, I couldn't move it to next one, but the the pi is some of the change in this PR is required for other patches.

C

For example, we shouldn't we shouldn't install some open flows in in some scenarios that they don't apply. For example, the net policy only mode and I will extract some code from this PR to and and make a new smaller one to fix that case and for the third one I think it's a it's a behavioral change of financial proxy to align with the cube Pelosi that when a service I would not has no endpoints on a note, it should be rejected instead of dropped.

C

So a single ways should try to merge this in this release, but it's not very critical so far and the first one. Let me just mentioned that I think it's ready to merge, because there is no real code change, but we need to ensure that, after the switch, everything should work. I should should not add, should not at least be worse. Then what do we have with endpoint API.

C

And the next one is a new future uh in addition to the The Seven Year policy feature, and with this PR we couldn't support um the seven nail policy logs if with the same flag in the National policy, spec enable logging and I think this Pi is already close to merge. We just had a few minor comments on the single Geo, always adjust them Timely.

C

There should be no list for this piano and for next one. This is to fix the the scenario that, when, when the client use TCP I use DNS over TCP protocol to resolve a domain, and previously it was not supported and leading to some where the issues when, when this happens- and it costs a lot of test, flake and I- think the current uh things is uh ready to merge. Even no other new comments is left for it and uh but currently we're waiting for the CI apparent to pass all validations.

C

And this the next one is also uh an enhancement for multi cluster scenario and I think it it has a workout channels between Gateway nodes or for each multi-cluster, each each cluster, and currently we we had. We still have some review comments on this PR and I think is still being discussed and being addressed.

C

We may need a few more days to finish the code review and the code change.

C

Nestle is a new test. I will first skip the test related PRS, as they are best to have to must have in this release and for the next one. um I think this has already got gotten enough reviews and has been upload.

C

um We, we are also waiting for the same paper plan to finish the validation of the SPI before it can be merged, and this one is to introduce a Rancho CI pipeline for enter and the right term regularly.

C

I think it has got a blue for some reviews, so.

D

C

And I would do another round of check before much of this one and if anyone else could take another look in webbing, much appreciated and the next one we have gotten to uploaders for this PR and I. Think it's ready too much as well and for this document Pi is. It had been merged and for more traffic, more support for Mad cluster giveaway.

C

Yesterday, I took a look at this behind the uh I left, a few comments, a single lines, working on.

D

The chest I'm still working on this one and I think we'll push the latest one today.

C

Sure, yeah and uh NASA is also new test. I. Remember here already got a blue or from Wayne, yes yesterday and perhaps it it gets another update after addressing some conflict and the next one. The last one is a unification of some processing in this package.

C

There should be no big change, but I. Remember last time, I reviewed it. uh There is some, then there are some um bugs and we we need to fix. Perhaps they are already adjust. I will need to take another look and any other reviews from other members are welcome.

C

I think that's all of the current status of this release. If anyone has any question, please let me know.

A

No question from me: so thanks uh thanks the channel a lot for the overview, so basically uh our new release date is a Wednesday, March 22nd and probably the cutoff date for merging PR since uh Monday next Monday. Probably the 20th. Is that correct? Yes, yes, cool, okay, just a very quick update for Tia! Instead, we are ready for the release and we don't have any outstanding patch. So that's all very, very, very, very, very quick update. So um right. So this is uh about Andrea 1.11. Is there any comment?

A

Any question for the team regarding Andrea 1.11.

A

Okay, it seems that we don't have any questions. So, let's move to open discussion, uh I don't know if there is any any topic that you would like to bring up, but let's keep the meeting open uh for a minute or two waiting for topics or for anyone to bring up.

E

um Actually, I want to propose some actions for the future PRS related to Windows tests. um The the truth is that several weeks ago were observed, Windows end and test failure in rpr's. After some investigation, I found that the root cause was not was related, one with one case, which was not supposed to be run uh even to be spotted on Windows. After checking the community history, we found that the pr was not verified on windows before it was merged.

E

um The the fact is. This is not the first time movies which we have seen Windows test failure by a non-related patches, so I want to propose some actions for all the contributors in chair.

E

um We expect that the patch owners try to pass our Windows tests on your on your PR before it could be merged. If the patch has some um candidate impact on the Windows logic, one easier way to check if the battery may impact Windows is you could check if or not a file with a suffix underscore windows.go exists in the in the folder, where your patch has has been covered, and a second action is that so we will expect all the contributors to pass Windows containerdy end-to-end test before the patch can be merged.

E

It is a. It is a mandatory requirements so that if the the patch has some some benefit impact, we can find the the issue as soon as possible, and a third action is that to expect an interval for CI job could be into produced to run all windows tests and delete is the main branch, for example, a two-day check in their job.

E

In this way, if we can find the failures in membranch, so we can reduce some stress just before the release date and the fourth action is that so we expect Windows CI to introduce some enhancements to support learning more and more windows tests. I have created a easel to to track the CI pass. The CIA enhancements may include optimized, Windows CS scripts to shorten the time per job and add more windows test bats to support more jobs running in parallel. This is what I want to share thanks.

A

Thanks Wayne, so these makes complete sense and it seems that it's actually easy to break windows, and we should not be doing this and Easy in terms of additional tests that will be running for every for every PR. We only have the windows end-to-end tests, which are not very resource intensive or time consuming. Is that correct.

E

Yeah Windows uh Community end-to-end test is mandatory to be required to to run on every purchase, so if possible, we need a required flag in the test case so that it can remind the the patch authors and the maintenance to see if the past has has worked good.

A

That makes sense. That makes sense. No, my only question was uh this: this end-to-end tested does not require a lot of time to run or does not require particular amount of resources. Is that correct.

E

Do not take a most.

A

C

Perfect I feel, let's probably add the the the required mandatory one to to be triggered by the test, oh command, so that we don't have to type all the required test uh when people formation appear- and perhaps it also the the they also need to be a new skip windows and neutron test, uh build to skip that or otherwise we will have to run some um run this on that that uh LPS, which is absolutely not related to Windows, for example, a simple unit tested things or simple of this specific to Linux, just like the the other skip, builds for performance, nail, polish and and to end actually I tried to add the windows e2e to to make it uh mandatory.

C

But I found that we don't have this skip field. So I had to remove it too much of a lot of PRS.

D

A

A

Okay, anything else for today.

A

Okay, it seems that uh it's all for today and uh and therefore I wish everyone a great day great afternoon or a good night to uh folks that we joined for from the west coast and our next meeting will go back as usual on Tuesday on uh March the 28th. So thanks everyone for joining and talk to you in two weeks time.

D

A

Thank you, bye, bye, bye,.