 We'll give it like a one or two more minutes just to give people a chance to show up Can you hear me now? Yeah Yeah, I mean Usually works best if you bring your own laptop and Yeah Yeah, summer of code Public service announcements so summer of code is happening this year as every year and For students they can submit proposals and what would be nice is if we can identify some mentors or folks in the community who can Who'd be interested in actually working with students this year? Any volunteers and I'm sure to listen to the response there is You're very funny facial expressions, okay When I take it that nobody is please do get us in contact with only maintainers or reach out to myself or Chris if you're interested in Mentoring this year Okay That's done should on boy have a certification program Does anyone have more context on this I think we're gonna have to punt this then because I have no idea what this actually means Little bit context of this is like to get the certification program that allow the different providers Like Google have managed to Have Microsoft have and make sure they have so it's sort of like the Kubernetes Certificate program to make sure that they can interpret correctly Isn't like a personal certification program this is like it's a conformance test right This is what you're asking for basically, right Right, I think that that was what Chris meant on the issue Yeah, we probably should have something like that I mean, it's in that we probably should have that for both management servers and envoy instances Because these are both things which we claim compatibility between rights. Yeah. Yeah. Yeah. Yeah I don't know who's gonna drive that should we just do this next week and hopefully Chris and or Yeah, right Okay Okay, Todd do you want to present your status on caching? Yeah, I just wanted to do a quick note to let people know that I've been working on a catch filter and with a plug-in API that to support a variety of Whatever kind of cash back end you need and that I will be putting out a doc shortly so If you're interested in caching and envoy look for that one of the particular if you were getting feet wanting feedback in wanting feedback on is Things like a synchrony, you know whether caches need to work on multiple threads or just RPCs and bounce back to one thread But so sorry, it's a little bit incoherent because the document isn't quite ready I will be putting that out soon and I just wanted to let everybody know that it's coming cool Hey question for Todd. Yes. Is this based on the work that solo did with their caching filter or is this something net new? this is based on experience in Google so with Several different implementations and what we've done cool because I know the solo guys that Implemented some sort of cache into envoy already. I don't think it's open source though All right. Well, hopefully they'll see my the proposal when I put it out. Cool. I reached out. This is Josh, by the way I've been working a bit on the envoy caching stuff and also I think I met those That team at envoy con and I reached out to them if But didn't hear back to see if they wanted to collaborate If you have Incited into them, we'd be happy to chat with them. Yeah, I'll do that. I'll I can ping Edit and also Christian posture. Just join solo as well So I know a couple of people there all ping them and connect them with you Josh and talk Thank you Okay Next topic See I performance and I put down see I fires What Lisa and you know what the status is of that see I to merge was right now I think we have a pretty active fire, which is Most PRs or a large number of pay hours are experiencing our failures and yeah so I think last night I updated the GCC 7 PR and I submitted two empty commit and The sims is now reliable practice coverage and release with that one So but by like the outcome is like I limited the CPU Instance foot to four for coverage and five for release That will need longer see I time but like I think it's more reliable passing now So there's like two factor Makes this happen happen. It's like one is GCC 7 uses something uses more memory and Another is Google test is causing GCC to use more memory That's right To test is behind that. Okay. Yeah, since it's only causes test compile very slow and actually LLVM doesn't have this program a problem. So The ASN T send So the rest of tests like current tidy compile time options are running pretty well and they finished like in about 30 minutes Compared to coverage and release Yeah, I mean coverage is inherently slow anyway, but it's doing coverage things. But yeah I don't know what the best way to move forward is this issue Do you think separating out the resources from the GCC seven change would Make a difference or is it? Sort of you may as well just try get figure it out in that PR So I think now that PR the resources relatively conservative I'm not sure whether circles here. I have a problem or like there's like With their schedule recently since like I I do like have Open several GC instance and run darker with the memory limitation and I could run that with like five or six But like on circles here. It seems it's Failed with a lot of memory so that we might want to ask them as well Yeah, I mean We could probably get a more detailed trace of what's going on when we get these Yeah, yeah, I think I think after I mitigate this I'm planning to put the Some Profiling Time trace on that as well to get the art to get them into artifacts so we can look at them Yeah, that would be super useful actually just being able to look at the elements and Yes So I think one one like open question is kind of also gonna but not really is How much do we want to? Migrate to our VM since that will make CI performance actually pretty much better. I Mean at the other day since we want to support GCC at least one of our builds is going to be GCC based, right? Right. I think Mike previously mentioned that we can do a one GCC building only building a binary and test is all going through our LVM I'm a little more skeptical about that. I mean there are These are compiler bugs and so on and yeah implementation which make a difference and can show up surfacing tests So I feel it would be good to run all tests and I mean at least today if we I think we've released a clang But you know for practical reasons because we don't have native clang support yet for coverage It might make sense to Stick with GCC there and that would give us some coverage still of that Okay, yeah, and I don't think we can really claim that we're supporting GCC compile the tests and run them with GCC right I I Understand that part and I feel that too like So like we have we we need sort of work around to Make the CI faster since like This one is causing like about 90 one and a half hour for the CI to pass Which is I mean the other thing we can do is we can try get some GCP resources and set up one of these Executions of the CI pipeline on GCP. This is what Sri Ram suggested. He said it's pretty Easy to work with and you just need to give it some credentials and I mean if it's really as easy as that I'm sure we can yeah, I know one another work around that I actually I haven't tried but like I think we can do it is like use circles to high to just like create a VM on GCP with like 32 cores or something I can just build on that and kill them I think that's the way to go like the idea of you know, the other thing similar to that That came up was using remote build executors, but considering we don't even have remote caching stable I think it's probably premature to consider using RBE Yeah, probably just use the gcloud SDK to open a VM and run everything on the big VM like 32 Should be faster Yeah, yeah, yeah, well, let's see. Let's see what your PR does and I Guess the question is who wants to are you interested in owning this? So we'll do or someone else when I own this so Sure, I can take I can take that part. Yeah Moving yeah, sure Yeah, yeah Chris pointed out that it is an adventure problem with remote executors. Okay Yeah, I'll be probably too early for to Punch that into CI Yeah, please mess Okay Let's let's hope this all stabilizes today then cool Instruction notes, nope Deplication stuff You want to describe it Yeah, I think it's just worth noting at the community meeting that Essentially, we're pumping up the time for deprecation So right now when we remove when we Change a feature an envoy or move config around or something We can remove it the next release cycle, which literally we've had a case or something removed it two weeks later Right, they checked it in PR. We cut the release. They removed the code Which is a little bit fast for larger companies So we're just popping it up one more release cycle so instead of being zero to three months It'll be three to six and then for complete removals like hey, we're getting rid of NPN because ALPN is the new hotness, you know, major major things that make sense to kill off over time Trying to give it a year And again, if anyone has concerns comment on the PR and do it soon because I'm a holiday bridge today cool So, yeah, that's all from the agenda anything from colors Okay, I think we're done then see folks next time