Antrea Community, 21 Dec 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Antrea Community Meeting 12/21/2020

Description

Antrea Community Meeting, December 21st 2020

A

If you see the zoom voice, reminding us that uh this meeting is being recorded so for today we have uh stephen presenting retrospective for the andrea project for the past six months from the kubernetes channel, I I can see that he has prepared quite a lot of stuff, so this is going to be very interesting.

A

So thanks steven for preparing this and the floor is yours.

B

Okay, just a minute, I'm going to model this for those of you who are here back on june 15th. We did this once before with cody moderating, so I'm going to use the same tool. He did it's actually the vendor of that tool, renamed it. So it looks like it's something different. They renamed it to easy retro, but it's the same thing that was used before uh it's my first time using it.

B

So hopefully uh this will go error-free, but we'll see, let me post a link to this tool in the chat and it's also in the zoom. If anybody finds that more convenient than the chat.

B

And then let me share my.

B

B

Okay, because I'm moderating, I've got two versions of this open, but I think you should be able to see the standard one that you get yourself if you open it, and what we're going to do here is have people initially set uh proposals uh in these four columns or categories uh we'll allow about four minutes to do that, and this is entirely anonymous when we do these retrospectives.

B

The point of keeping this anonymous is that sometimes people feel more free to make comments in an anonymous form because doing it openly might feel like you're, criticizing the management of a project or the way things are going. So we're just going to keep this anonymous.

B

You should be able to open that link and participate in this with no registration, no email open an incognito in a browser. If you like, I think that all works, so the initial stage is that down these columns, you can add as many comments or cards as you like then, after this process, I'm going to hit this button that says, show cards and they will be visible and then we'll vote on them and just like the initial composition of the cards, there will be voting.

B

Let me see how many people do we have today.

B

uh They recommend that you have six votes, but if we have very few participants, maybe I'll cut that down, but it looks like six should be quite workable, so we'll do it that way, so um I think I've enabled uh submitting cards so just go for it hit the plus sign, I believe, and then you can add a card just if you want to nominate something that you think is working well, uh that hey, let's not change this or maybe even let's expand on it.

B

Put that's what belongs there, things that we want to stop change or improve our uh ideas for things that either need to improve, to make this project better, were so bad or such a waste of time. That maybe we should just stop doing them all together.

B

New ideas is just totally open-ended if you have any wild and crazy ideas, or maybe some that aren't wild and crazy, but things that can make the andrea project more successful. That's what goes there.

B

Then it's been six months since we have the last one and on open source projects, it's always good to recognize people or groups, or even ideas or practices that really worked out well to shine the spotlight on them, because by getting this kind of recognition, we can all see what's working very well for the group and amplify that and make it work better. So I'm going to allow four minutes to compose cards.

B

Can somebody just as soon as somebody tries one if you could just go on your microphone and let me know it's working for you and in the meantime, let me see I'll just play some background. Music.

B

Okay looks like we've got a couple added, so I think it is working.

B

This is a classical music piece called claire, dulune or moonlight. It lasts four minutes and when it's over we'll.

B

B

B

C

C

C

C

C

B

And we've got about 30 seconds left if anybody needs any more time speak up and we can lengthen this, but it looks like we've got a lot of cards already.

B

Okay, the music's over it. I didn't hear anybody speak up as to needing more time. So now I'm going to show the cards we'll go through these and read them and just make sure there are duplicates that we could perhaps combine and then we'll have a voting stage.

B

So, what's working well, the planned release cadence um the ability to manage the complexity of the project and keep introducing new features at a constant pace. More automated verification and tests were added to prevent common issues and guarantee quality really fast response. When a new issue is open, quick turnaround when fixing bugs good attendance and participation in the community meeting feedback on the kubernetes slack channel are helpful.

B

We should find ways to reach out to more folks.

B

So that is maybe a comment reaching out to more folks is maybe under the new idea, but we'll just leave it here. uh Monitoring and troubleshooting tools have been enhanced and the uh entreat continuous integration is very useful in getting better and better okay. So that's great, I understand them all. I don't think there's really any duplicates. So that's that's a good set of cards, things to stop change or improve.

B

We need to improve code review and we need more reviewers, more transparency and automation for the website generation, so we can accept contributions to the website. In other words, if somebody detects a dead link, they should have a way to fix it by opening a pr directly, more transparency, access to logs for some of the end-to-end test. Cases for windows and ipv6 fix the cla validation bot.

B

Many large prs got merged in the end of the release cycle, which conflict with each other and cause extra effort to rebase and potential issues need to improve upstream involvement.

B

Okay, the I think that those also are without duplicates- and I understand them so they're pretty well written, so I think we should be able to vote on them understand the objective here isn't to fix all these. It's just to prioritize these, and the fix is a different project process.

B

You know some of these fixes require individual attention or a long period of time, so we're not going to expect that we fix any of these at this meeting, we'll just prioritize what we want to do and maybe identify what we don't want to do under new ideas or activities. We should try uh layer, seven policy and visibility uh have a podcast video or something like that.

B

Bi-Week wheeler monthly uh reach out to major kubernetes distributions, tops k3s to see if we can work with their maintainers to provide first-class built-in, andrea support, encourage entree users to come to meetings to give presentations about how they use entree and what they like, and dislike.

B

Ingress and egress forwarding policies and office hours community meeting to try and reach more potential contributors. Okay- and I gather that's different from the podcast it should be. I guess- because office hours could be written questions and more of a deep dive- chaotic testing, for instance, something like chaos, engineering. So I'm imagining that would be a test suite that intentionally injects failures, just to see what happens um so yeah.

B

That's that's interesting and I think I understand uh for recognition uh of groups, people and actions, um I'm not sure, I'm great at pronouncing everyone's name here. So forgive me but uh say chang for his hard work, maintaining the andrea, ci and continuous improvements. He makes everybody in the team involved in upstream kubernetes work who gives back to the community while getting entrea some name recognition.

B

uh The team that worked on ipv6 for andrea uh quan, for handling, challenging technical problems, antonin for managing release and talks and abhishek for all the k, kubernetes upstream work. Okay, it all looks good. I think they're all well written we're going to move to a phase now where everybody is going to get six votes, so you might as well use them all.

B

I don't honestly know whether you can combine them and put all six of your votes on one or not, but whatever it is, it is so I'm going to press the button, I see as admin and enable voting and go for it. I believe you click those.

B

Finger of icons and we should see the numbers going up.

B

I'll start the music again and maybe we'll give this stage one or two minutes I'll just wait until I don't see things going up.

B

C

C

B

B

Okay, I think the upvoting is quieting down, but I'll give it 30 more.

C

B

Okay, if no one needs more time or if you do speak up right now, otherwise we're going to call the voting to a close.

B

And I'll take silence as an indication that nobody needs more time. Can everybody see those votes? I mean I have a special screen, I think as admin. So I don't honestly know whether they're exposed to everyone yeah. We can see them yeah, okay, great!

B

So, let's just go down the first column and see what the the leading things are. Well, first of all, pretty clearly the entry ci at the bottom, one in the column and the green column, uh got a lot of votes with eight. There was no, but nothing else even got half that um so yeah, let's pat ourselves on the back for that, um and then the comment is that it's getting better and better.

B

So I think we'll leave that alone, but everybody recognizes that, hopefully, the people responsible for doing that are over in the recognition column too, and I I actually looks like I see that with uh chang, um then the other things leading with threes, three votes, the planned release. Cadence, uh that's interesting, because I went back and watched the video from june and I think there were some people thinking that perhaps things were uh moving too fast.

B

You know, there's always a trade-off in a software project where you can make it schedule driven feature driven or quality driven, but you can't do all three of those at the same time um push too fast and something else might have to go where you're getting more releases, but either less features or poor quality. But it seems like there's a fairly good vote for the pace being about right and I didn't see anybody nominate over in things to stop or change anybody asking for it to slow down.

B

So I think that I'll take that as an indication from the group here tonight that uh generally people are okay with the way things have been going, um then the final one, with three votes: the ability to manage the complexity of the project and keep new features at a constant pace. It sounds like both the feature process and the pace of introducing new features is okay with people.

B

Then I'm not going to talk about the ones that got zeros.

B

I guess that might even apply imply that the person who wrote it thought some of the other things that popped on the board were better choices and we have limited time, but the people appreciate the really fast response time when new issues are open, I'm sure the users appreciate it more than the people here. I don't know if we have any users on this call tonight or if it's all devs um any users here, uh please speak up. I'm just curious.

B

Okay, I'll take silence as no users this evening, um then the monitoring troubleshooting tools have been enhanced. Oh okay, scott is a user, so we do have one user- okay, well, that's great, um hopefully, by the next review that uh we'll aspire to having in either one quarter or six months like the past cycle will have even more users because really with any open source project. That's the whole objective.

B

Let's move on to the things to stop change or improve.

B

We have a clear leader there with seven votes. Many large prs got merged at the end of the release cycle and they may conflict with each other and cause extra effort to rebase and have potential issues.

B

You know this isn't unique: andrea, wouldn't be the first open source project or even closed source project to have big rushes for the door at the end of a release cycle, I'm not sure, there's a perfect answer here in terms of um you know a process for merging npr's. um You could, you know you might aspire to. First come first serve, but the trouble is that they're not all created equally in complexity.

B

Some of them might obviously be really low risk and only require a simple review like fixing a typo so often things that came in later than somebody else's pr jump the queue.

B

B

I have a great solution with that for that the things you might have control over is whether project admins help with the rebase or you push that back on the person submitting the original pr to you know, do the rebase and as a retrospective, I don't think you know we're not going to solve these identified issues on the call here in the meeting, but we should recognize that people perceive this as troublesome uh and see if anybody's got ideas for how this can be resolved.

B

um I've seen in some other kubernetes sigs for the kubernetes project itself that there is a process I've seen done in the storage sig that has cutoffs for prs now, there's a downside there and that it might result in prs missing the train for the next release and having to delay an entire release cycle. But one way to simplify this is to have hard cut offs so that you can't have a new one appearing.

B

You know within the next within the last x days of the planned release, uh but there's an obvious downside. Once again, uh it's not even my role to fix this, but let's put it- uh I don't, I think, maybe we should discuss this one further in the slack channel or some other venue that is appropriate for people to submit proposals for doing something about this and have others evaluate the proposals and offer opinions on whether to do the change or not.

B

Okay. Moving on the next leading vote uh category was improved code reviews and we need more reviewers, um okay, I'll I'll. Take that as that with five votes uh that probably shouldn't be in dispute. um What can you do to get more reviewers? Well, um you know, reviewer, reviewing is often thankless work, so maybe at a minimum um over in the recognition column, if anybody really stepped up and did a lot of reviews, recognizing people for that kind of work would be a great idea.

B

The you know it depends on the reviews too, where some of these might be fairly simple and others.

B

You don't want to merge a high-risk pr without review by a senior architect, but maybe there's a mechanism where we could put in a triage system where someone could do an initial screening and maybe have a volunteer tier for low-risk reviews to just you know, check check them out and delegate to a bigger team and it's a good way to even coach, uh less senior reviewers to move up to higher levels, uh just throwing that out as an idea.

B

But really we should discuss ideas for that, uh maybe over in the community slack channel too, and see what we might be able to do about that.

B

B

Card in this column, with three votes: more transparency, access to logs for some of the mdn test cases, I'm not familiar with that myself, but it sounds like these test cases could either be an individual developer's unit test before they submit a pr or something that takes place after a pr gets submitted, I'm not sure which, but um I don't know.

B

I I'd- ask whoever wrote this to maybe clarify what the category is for these logs, so that we uh at least for me personally- it isn't clear what the next step would be and once again we could I'll run on jenkins, okay, okay, so that that actually is triggered by the github system.

B

Okay, thanks for clair clarifying that so, um okay and once again, I'm not that familiar with jenkins and what you need to do to get access to those logs easily, but if the system behind it itself could be improved, that would seem like a good idea.

B

uh The next uh vote getter with two more transparency and automation for the website generation. Okay, so I know I've spotted some of these dead links on the website myself.

B

So I think this opening a pr directly, I gather, would be a proposal that somebody could go at the website source if you will and submit the fix directly. So I don't see much downside to putting that in place, but point taken- and at least two people are in favor of that.

B

uh Let's I see we I I planned on this being 30 minutes and I'm already at 30., so I'm going to speed it up we'll move over to the third column.

D

Stephen sorry to interrupt, I don't think we have anything on else on the schedule on the agenda. So salvatore can correct me, but I think you can definitely take 15 minutes more yeah.

B

That's correct, that's correct! Please go ahead!

B

All right and I see now I I glanced over at the chat and see that somebody made the comment that windows and ip6 are tested privately, so um the logs can't be exposed yeah that that seems like a great idea to somehow do that, because I'm even guessing that we might have contributors who don't have a test bed to even get to windows and ipv6, so uh getting those out there invisible seems like a fantastic idea.

B

Maybe it's even if there's a system behind copying these files to some public uh share where everyone can get them, uh probably with a timeout, so they don't build up in perpetuity, um okay and somebody's suggesting using vpc. Well, once again, let's cut and paste these comments on how to fix it over into the community slack channel, because I don't think we're going to have the time to come up with a group consensus on what the fix should be. But these ideas are great.

B

So back to the new ideas column, it looks like the lead, vocator lev layer, seven policy and visibility. uh That would be uh new features. um You know that I can't argue with that, um especially if uh we have users voting for it. um The other one, with card with six votes reach out to the major kubernetes distributions uh cops of okay. That wasn't cops before so thanks. I understand what that is, k3s see if we can work with their maintainers to provide built-in andrea support once again six votes.

B

I think that that's about as high as we're getting here in voting for these and not much dispute here.

B

Once again, it's not the place to solve this, but I will make a remark that I'm familiar with at least one which is microcase, and I think this is true of a lot of them, where you ordinarily, don't have a situation where you expect the distribution maker to do the work.

B

They likely have a pr process where what you do is reach out to the distribution and ask if they have a plug-in system for cni's, maybe even ask them of all the cni's they support now, which one is the best implementation, because you want to model yourself after the best one, not not a random one.

B

So my experience with micro cates in I I went to a micro, cates uh community event and just asked them, and the leader of that group was very amenable to helping out even walk through the process of a plug-in showing what's involved, by opening up the source line by line. So I think that these distribution backers are very much in favor.

B

My suspicion is that the mid-tier ones are actually more in favor of it than the shall we say the leaders just because they're often in a position, just like the andrea project of trying to establish some momentum.

B

So it's a win-win situation for both of us, potentially where the distribution vendor gets external contribution and momentum, and then we submitting a cni plug-in would be getting it too. So the low-lying fruit with maybe much more vigorous cooperation might not might be kind of the mid-tier distributions rather than the more dominant ones. But of course the dominant ones have uh perhaps a lot more user adoption.

B

uh Moving on uh we'll move down to f the card with five votes that office hour. Community meeting to try and reach out to more potential contributors sounds like a great idea to try I'll leave that to project management. To try to that.

B

You know, aside from declaring the office hours, and I think you'd want at least a few people to show up maybe go out there and promote the idea that we'd like to have one and the first community member that would like to ask questions: let's steer them into an office hours event and uh just schedule it.

B

It probably isn't a bad idea to have that event be either a zoom or, if it's on slack for time zone reasons in just q, a a q, a stream to preserve this, so that we'd build up kind of a frequently asked questions thing uh that is preserved and does you know just as a one-time event that goes away um moving down to four votes, encourage entry users to come to meetings, so we've got scott tonight, uh but somebody's asking for a presentation on how they use it. What they like and dislike?

B

Well, I hate to put you on. I won't put you under a spotlight and force you to answer tonight, scott, but in the future.

B

If you'd like to volunteer to do this, apparently it would be quite appreciated and of course it would be even better if we can find a few other users, um then finally, we've got a vote for having a podcast, video, bi-weekly or monthly.

B

um You know, I guess I shouldn't take this over, but there's an andrea, dedicated podcast that we run ourselves, but there are also podcasts like the kubernetes one out there that already have a viewer base. uh It strikes me that perhaps- and whoever put this card didn't say whether it's an entree owned, podcast or a more generic one, but I'd suggest that at early stages, when we don't have this huge user base and viewer base, maybe if we could crash the gates and get on podcasts that already have a viewer base and are popular.

B

We get more momentum, um just throwing that out there, but we'll take it as an action item to consider that we should have a venue, a podcast, video or something like this uh and discuss later, where that is. And finally, let's move over to this fourth column recognition- and this is perhaps the most important one- we've got three of these with seven votes, so the first one will just go from top to bottom.

B

As I see them, everybody in the team involved with upstream kubernetes work who gives back to the community while giving andreas some recognition. So I think this is people recognizing that being active in an open source project beyond andrea, specifically, kubernetes is a big factor in getting andrea, but not just name recognition, but adoption that you know. We can't just stick here in these meetings that we conduct ourselves and have every all the great work here going on underneath a barrel that nobody ever gets to see. So it's this crossover into the kubernetes community.

B

That is really going to be the thing that moves the needle and it gets actual adoption. That makes a difference. um So it's great to see this, I don't know how to call out the specific names of people other than um you know. Good work, everybody, let's move on to the other one with seven votes, uh quan for handling challenging technical problems.

B

So I see you on the call tonight. Congratulations, uh you know. If I had this, my own department of open source under the map of you under normal times would have had a swag uh budget, but we should see what we can do for these people. Getting these seven vote clouds to get them some actual tangible recognition, if nothing else other than you know a t-shirt or something that they can wear proudly.

B

So maybe somebody in a position here to kind of get some kind of tangible physical recognition for these people would be a good idea and then antonin also got seven votes for managing releases and talks. So um I'll join everyone else here in applauding uh antonin. For uh this- and it looks you know with with seven votes, it seems pretty clear that uh he's leaving an impression on the team uh four votes.

B

um You know that's still good uh abhishek for all the upstream kubernetes work and I guess in a way he deserves, since the give backs and participation in the kubernetes community got seven votes as a general category. If he's getting four votes for being in that category, maybe he should be deemed another.

B

Seven vote getter by you know, by being a big part of the seven votes for the kubernetes upstream community work, but uh congratulations, abhishek great work uh keep up the good work, then uh chang for his hard work, maintaining the ci uh and the continuous pro improvements he makes to the ci uh system. Now he only got three votes, but I'm I'm.

B

I have to say that if you look over at the bottom of the green column, one somehow the ci system got eight votes, so maybe he's a saying: people only got to allocate six, but maybe if they had gotten, seven he'd be up there at the top two and he owns a piece of those eight votes down at the bottom of the green column. As I see it, so hey great work and I think everybody appreciates it. um Then um will the final one here: the team that worked on ips ipv6 support for andrea.

B

So um that's recognition for everybody across the group. It means you know it's always good, to see a team that likes working together to get things done, and that's what this looks like to me that everybody yeah it's only two votes, but I think people might have naturally wanted to put put their votes behind some of the other categories, uh but this one is great work by the team. So thanks everybody, so that's it for the retrospective I'm going to take a a screenshot of this board to preserve it.

B

uh This easy retro thing is in the free category, uh so the vendor doesn't guarantee they preserve this in perpetuity, but I'll cut the screenshot of this retrospective put it in the slack channel and then whoever takes the notes and puts them up on github, perhaps can cut in, can go grab the screenshot picture out of the slack and go put it in the notes to preserve the outcome of this retrospective and um with that said, I think I'll uh call this retrospective to a close.

B

Now here I am running a retrospective for the entry of project itself, but given that this is only our second retrospective, perhaps since we have a little time left, we should have a retrospective of the retrospective.

B

Does anybody want to come, enable their mic and give comments on how you think this retrospective went and uh any ideas you've got for making the retrospective itself better.

A

Hello steven this is salvador. I think it was a very nice exercise. It allowed us to get a good feeling of of the current status of the project. I think that, um to be honest, the only thing that I felt that could be better in retrospective was having more votes for cards. uh There were a lot of proposals and maybe six votes weren't enough.

A

You should be mute, I guess.

B

Okay, that's that that's good feedback and we'll consider it for the next time I read about connect conducting these retrospectives and a lot of that six depends on how many people you have submitting cards where I perhaps prematurely came up with the six. To be honest, I only did it because when cody did it last time it was six and it seemed to go okay, so I went with six but uh yeah. I think we got quite a few more cards than we had in the last one back in june.

B

So if you've got more cards to vote on, perhaps allocating more votes is a great idea.

B

um What you don't want is so many that everything comes in tied, so there's kind of a happy ground here where you get enough to get separation, uh but you don't have people unable to recognize things that they thought were tremendously good or horribly bad.

A

B

Any remarks from anybody else.

D

I just think it was very nice to uh be able to compare it to what we did at the previous retrospective as well, and see that a lot of quote unquote negative feedback. The previous retrospective was about the ci system and how it was making the development process more complicated, and it's really uh pleasant to see that uh this time around like uh this year, is what gets uh the most uh cuddles here. In the first uh column,.

B

Yeah, I I think that that that really is why we want to have these retrospectives, because it it is the mechanism where we continuously improve. You know really.

B

One solo developer isn't normally empowered to do a whole lot, but as a group, we can take on larger things and also as a group. Maybe five people are observing some issue or some opportunity to improve things, but they've never made remarks to each other and it goes unnoticed. You know. Sometimes people don't realize that they're, not the only one seeing an opportunity. So uh it's good to see follow up on these and what I'd like to do with this board, like I say, I'll, cut and paste the screenshots summarizing those votes.

B

But the way to make things happen is to remember the ones that got upvoted in these categories and follow this up with plans to change things.

E

Actually, did we go back to the previous retrospective and compare as to how we have improved and hopefully not redressed from the previous uh items? Well,.

B

We didn't, but let me see if I can um here just a minute. I don't know if I have a capture of that retrospective on my computer. I might I'll go look for it if I don't I'll have to scroll back in slack all the way to june 15th.

B

Maybe while I look on my computer, because I think I took a screenshot back then, but could somebody else, maybe scrolling back in slack to june takes forever? So maybe if I can get a volunteer to go, try to do that. Well, I look to see if I have a screenshot on my computer.

E

Maybe not today, but probably okay, sometimes yeah.

B

Yeah, okay, yeah because I'm not seeing it so it would take a while. But I'll tell you what I'll upload the screenshot of the voting from this retrospective and I'll go find the one from the last one and I'll upload that into the slack channel too.

B

We can see if we uh erroneously ignored something that was called to attention back then, but I think antonin's memory is correct, that you know some of the things that came. I remember the the the ci being raised as an issue. I think there were some opinions about the pace too, but they didn't seem to come back out this evening. So maybe those got resolved too.

B

So I've got action items to put a couple of voting summaries on the slack channel. The team has action items to do something about the things that got upvoted. Does anybody else and uh a long-term action item is to evaluate the number of votes in our next retrospective?

B

uh Anybody else have observations or comments.

B

Okay, I'll take silence as no and then I'll turn it back to the community meeting moderator. I think. That's um you uh salvatore yeah. If you.

A

B

To close it or bring up something else, go for it.

A

Yes, uh yes, I have uh the moderator role here. Thank you very much steven uh for presenting this retrospective and I don't know if there are any other topics, but uh I is there any other question on the retrospective any sort of genetic question that you might have any specific question about. The topics discussed in.

A

Retrospective I'll wait 10 seconds as usual for an answer.

A

All right so, as we have still nine minutes left in the meeting nine minutes, 25 seconds to be precise, uh we don't have any other topic in the agenda. So if there is anything that you would like to bring up both for uh features that are currently planned for the next century release or for general community topics, please feel free to.

A

A

And therefore it seems that uh it is all for today this is uh the last meeting of 2020 and uh for next year uh you know that has been again a discussion of uh rescheduling the meeting, as the current timing is particularly difficult for for people in the united states, especially for those not living on the west coast, uh as, as you know, the meeting falls on midnight for for the eastern coast.

A

So, as you know, from the kubernetes from the slack channel uh for this project, we are either discussing the option of having an alternating uh meeting for andrea, one in uh us, friendly time zone and one in a uh pacific friendly time zone, or we are considering as reviewed in the in this retrospective. The idea of adding a second office hours meeting, which will have will, however, have a slightly different focus than the community meeting.

A

It will be more to address questions for the community, uh discuss uh general topics about the project rather than discussing the mothers pertaining to upcoming and upcoming releases and futures being features being planned. So that is all that I had to say for, uh for this instance of the meeting. So the only thing that uh is left to do is wish everyone, a great and uh to this year a relatively great end to this year and uh see you see everyone in 2021.

B

Okay, see you next year.

D

Thank you happy holidays guys. Thank.

A

E