All Core Devs - Testing #51

Transcript

Mario Vega: Hi everyone, let's just give a couple more minutes, until more people join, and we can get started.

I think we should be re- ready to start.

Pooja Ranjan: Real life.

Mario Vega: Thank you so much. Good day, everyone. Welcome to ACDT number 51. Today is September the 1st. Yeah, let's just get into it. The first item in the agenda is the Fusaka DevNet status update. I assume, since there's no launch yet, I would like to focus on DevNet 3. A status, since that was the main issue last week. Do we have, anyone from the Eat BandOps team wants to chime in, regarding DevNet 3?

pk910: Yep, I can give a quick update. So, we had that unfinology test on Definite 3, We fortunately recovered from that. But right now, we seem to struggle with finality again. I'm not sure what's going on there. Yeah, question is, is are there any outstanding, open points from client teams that are Pending right now, or…

Emmanuel Nalepa: I can speak for PRISM. So Prism has… three downs nodes, one with the reconstitia. For RES and Aragon, we have both issues with execution payload, which are invalid from Puiz's point of view. Yes, basically, not matching routes and… And, unvalid block. And so, we are, checking… with Russ and everyone. And from Nimbusuel, I don't know, I didn't check. It's probably a memory, issue with Nimbusiel. Yeah, that's it for Prism.

Pawan Dhananjay: I think for Lighthouse, I… I haven't checked the… checked it since the weekend. I'll check it out today, but We have made progress on a new sync, sync algorithm, and I think we have ironed out almost all the edge cases. We have one or two, more edge cases that we need to fix, but we know what they are. So, we should have, like, a like, we should probably be merging the new sync strategy onto our Unstable soon, and Yeah, I will look into the notes that are offline right now, I haven't looked into those yet.

Matthew Keil: For Lodestar, we've made a lot of updates and progress towards our syncing, and things are looking very good. All of our nodes are synced. We just have a ref node that, is optimistic, but, we'll look into that and try to get that sorted.

Mario Vega: Any other clients that want to chime in? Is Teku on the call? Don't want to give you an update? Deku or Nimbus? Alright, okay, that's good to hear the syncing issues are dissipating. Is there any other tests that we want to do, then, in DevNet 3, or is it gonna be deprecated? Also, is there any plan on the dates for DevNet 5 so far?

Matthew Keil: I had a quick question on DevNet 3. It seems like, we were doing really well up until middle of last week, after we recovered from the forking and non-finality testing, and it seems like things are, just as a network, have kind of degraded. Was there something that triggered that? Like, did you guys run another test, or, like… Something that… that caused it, so we could do some root cause analysis.

pk910: I think besides of the initial, Plant? back, there was no interruption since then. We've just tried syncing them. And I'm also not sure what's happening right now, because he's lost finality again. Half an hour ago.

Pawan Dhananjay: I have another request for DevNet 3. Like, once we've recovered from the current non-finality, like, before we shut down the DevNet, like, can we have, like, a planned the similar, full chain split kind of thing that you guys did last week, so that… and, like, Like, let us know beforehand when you're planning to do that, so that we can sort of monitor our nodes and See in real time, like, how… They're doing, like. Last time when you did that, like, we… Like, we went around to, like, check the notes while it was in the… Chain split stage, so… If you could give us a heads up before you do that, next time, that'd be really helpful, because,

Yeah, we could check the logs and metrics during the chain split, and those would actually give us a few helpful yeah. It would be helpful for us to look at them while That is happening.

pk910: That totally makes sense. We also plan to redo that, chain split test. Because, yeah, once everyone has fair… merged in, there are fixes, we want to retry it if it has any better… Positive effect or not. So yeah, we'll give her a heads up before we start next time.

Mario Vega: Hey, PK, so one question on that is, that is… the plan is to reproduce it on DevNet 5, is that… is that correct, or are you guys planning on reproducing it again in DevNet 3?

pk910: I think once definite 3 is fine again, we can also reproduce it on definite 3 again, just to see if it has… if the… Improvements had any positive effect or not. Definite 5 will be a lot bigger, it… Probably makes sense to add that test there, too. But, yeah. To be decided, I guess.

Mario Vega: Excellent. Thanks so much. Alright, any other, comments on DevNetri? As far as I understand, the plan right now is to keep, monitoring DevonTree until things get better. And we regain finality again, and then we can decide on Definite 5. Or do you guys have plans for this week for Devon5, or is that pending? PK, do you know?

pk910: It will probably be next week, when everyone has their releases ready. Or at least the fix is, matched in.

Mario Vega: Makes sense. Thank you. Alright, Yeah, thank you. So, the next topic was syncing issues, but I think Lodestar, your guys already, chimed in on the topic, and also Lighthouse. Is there any other findings regarding syncing, that any other teams, stumbled upon, or… Or got the opportunity to analyze on last week?

Matthew Keil: We did have some findings, but I actually, I agree with Pawan that when we do the non-finality, or before we shut down DevNet3, it'll be great to… To do that chain fork type test again, so that we can monitor it as it's going, because it'd be really good to see how, things diverge, and how our algorithm is actually picking it back up.

Mario Vega: Got it. Yeah, of course. Excellent, thanks. Any other comments regarding debit tree syncing? From clients… from the execution clients, was there any findings during the non-finality test that anybody wants to share? Alright, no comment, so I guess everything went fine there. Alright. Yeah, if there's no further comments on DevNet 5, I'm sorry, DevNet 3, we can jump into the next topic, and just to reiterate, DevNet 5 would be probably launching next week, so we should be good there, and we are gonna discuss that in the next week's call. Alright.

Great, the next, oh yeah, and one comment from PK is that… he's gonna monitor the three notes, if I understand correctly, in regards of the non-finality that is currently happening. And yeah, we're gonna redo the split chain again on Definitely. With a prior notice. Good. Alright, next topic is gas limit, testing updates, I think… Camille is the main person here, Do you want to keep an update on this?

Kamil Chodoła: Yeah, can you hear me?

Mario Vega: Perfect, yep.

Kamil Chodoła: Yeah, good. So, last week, we worked quite a lot on the… on making sure that we are able to test stateful scenarios on top of any existing Database, an existing chain. So, yeah, we have some successes here, so… basically, I'm able to use EEST Execute and start any EL client on top of any network, but in some isolated way, that I can run execute, consume all of the transactions which are being produced by EST, and wrap them as a new blocks, and basically record all of the usage, and record all of the new payloads. So, this opens a door for all of these stateful scenarios which we want to execute, want to try out, like Zen.

et cetera, and see how… how it performs, and we'll be able to execute that on top of BlotNet, as well, to see how integrates on higher… on bigger states. So, yes, right now, working with Johim on that. Johim is writing a new scenario in EST, which will be mainnet-specific. And, I will be working on the generation of tests. Also, like, the EST integration is there, I just need to polish a few things and make a documentation, and I will merge it from this week. So, by default, gas benchmarks will be utilizing the ESD scenarios. And yes, we'll not focus a lot on compute scenarios anymore, rather we'll start switching to the stateful scenarios, just to see how it works on top of manual.

And of course, I will execute all of the compute scenarios on maintenance data as well, just to see if this is getting… Much affected, but not expecting, like, a huge difference. So yeah, on that, and today I got some questions from Bezo team about how to run the EST test, because they had some issues with that. Hopefully, this is moving forward right now, so Bezo can check the… visiting team can check the current bottlenecks on their side.

Ameziane Hamlat: Yeah, I can talk about that issue. Thanks, Camille. So, Yeah, I could actually, once, specify in the good Genesis file to run all the tests. It took some time, but now I got all the results and could reproduce the issues. But yeah, we have been, we have someone working on… Improving, different arithmetic opcodes. Like, mod, mill mod, but, we, we just need to take some time to, to test them, because it's like, A new implementation, so we just need… To have good, good tests, And, and then we can go Fraud.

Mario Vega: Yeah, I think someone from your team reached out to see which tests they should be running. I think a combination of Ethereum tests and the latest release from EEST would be the sanest and safest way to go. We are also preparing a new release from EEST. It's probably gonna come out this week, and it should contain basically everything that you need to get full coverage if you want to fully test your new implementation. But yeah, we'll reach out to you guys when that happens.

Ameziane Hamlat: Okay, thanks.

Mario Vega: Cool. Anything else here on the topic that gas… of gas limit? Do we want to go into the discussion about 60 million at the moment?

Kamil Chodoła: Yeah, so, as we discussed probably last week and last weeks in general, $60 million can be considered only if you'll have fixes from Bezu, some decisions and fixes from GAF about, modest improvements before Fusaka. And before Fusaka will need to analyze then, and once we'll have some outputs from stateful testing, then we can make some further decision about 60. So, like, 3 points. Once we'll solve all of this, we can… Start considering rolling out 60.

Mario Vega: Excellent. Alright, yeah, cool. On the Modex issue from Geth, is there anything that you guys need from… Anyone? To proceed on this, or is… are we good, or…

Marius van der Wijden: There's nothing needed, we… slowly working our way through it. We have, PR that improves the gas computation, which should give us between, like, 2 and 10% improvement. And then the… the bigger thing is to… to actually have… Actually upstream our stuff to the… to the Go Standard library. This is just a longer process. This is a priority for me, so I will, I will, update next week. About concrete, the when and how we can move… forward with this.

Mario Vega: Sounds good. Thank you. Alright, yeah, that seems about it. Cool. So, should we move to the next topic? The next one is, Glamsterlam testing updates, specifically EPBS and local access lists. I see Tony on the call, Or anyone else who wants to give an update on Black Global Access List?

Toni Wahrstätter: Yeah, no real updates from my side. Maybe if Felipe is on the call, he might have more insights into how we are doing on the testing side. Or Rahul.

Felipe Selmo: Yeah, I can give a quick update, The spec side is cleaned up and passing CI. And the testing side is also talking to it. I think, because we're merging, Both of those codebases. It's kind of waiting for a little bit. But it is, it's pretty good to go right now.

Mario Vega: Awesome, thank you. Great. And on EPBS, is there anything that we have to discuss today? Dan, anybody wants to bring up?

Zoom user: Can you guys hear me?

Mario Vega: Yes, perfectly. Yep.

Zoom user: Yeah, so, this is Terry, sorry, my display name is messed up. I guess one thing to give an update on is that, like, we have been working on spec tests on the… on the consensus spec report, so that's on the, the new upgrade name, Glorus, I don't know how to pronounce it. But we have sped tests for SSD test vectors and stuff, and then all the block operations are there, and the EPOB processing. So, as you implement the new spec, you could also, like, add a spec test. There are a few bugs here and there, which I won't be specific on the bugs, but yeah, if you have any questions, just ping me. Thank you.

Mario Vega: Thank you. Alright, and… Talking about consensus specs, I forgot to mention there is a new release. I remember there was a consensus issue. talked about this last week on this call, so I assume that the new test vector is included in this new release. I'm gonna link it right now. Yeah, I don't have more info on this, but please check out the release notes, and just make sure that everything's there, unless there's anyone here to chime in on details of that. But I don't see anyone, to be honest. Anyway, just wanted to mention it. There's a new release, please check it out, for Consumer Specs. Cool. Alright, and the last topic I have here is Sunnyside Labs. Is there anything that anyone from Sunnyside wants to chime in?

If not, that's okay. Yeah. Okay.

J Sunnyside Labs: Yeah, hi, nothing really particular. This week, we will be starting to prepare for the DevNet 5, if we were going to launch it early next week. And continue testing, in between.

Mario Vega: Okay, sounds good. Cool. That was everything in the agenda today. Is there any topic that needs to be raised anybody wants to talk about? Otherwise, I think the priority right now is trying to fix the finality situation in .NET3, and the bug further syncing. If possible, we're gonna retry to do the finalization test once we have FinalLG again, and then we can talk next week about DevNet 5. If that's okay. Alright, yeah, I think that's everything for this week, quick call. Thanks, everyone. for joining. Okay.

Marius van der Wijden: Thanks, buddy.

Tim Beiko: Thank you!

Mario Vega: But…

Toni Wahrstätter: Thank you.

Transcript adapted from ethereum/pm, CC BY-SA 3.0.