Dencun Interop Testing #38

Transcript

[Music] J [Applause]

h for

stream we are live uh for this testing call number 38 um so don't have too much on the agenda I guess having an update on how the spamming is going on dev Nets uh would be a good start then if there's any concern thoughts about client teams on anything um and I think finally would probably make sense to just touch on like the call schedule we want to keep for these the next one is right between the two test Networks so it seems reasonable to keep it but um yeah I want to open it up if people may want to make this a monthly thing or change the Cadence as we sort of wrap up the fork um but we'll also be starting a new one

so maybe we stay on this two we Cadence um but yeah per Barnabas do you want to kick us off with some definite spamming updates yeah um I had shared the initial girly analysis a couple days back um andon um so just a follow up on that one most of the stats look pretty much exactly the same um there's no real difference so I think whatever takeaway people have from that post still holds through like over the weekend um but in general if someone wants to poke around we still have the data here uh we haven't changed much about our strategy um The Blob spam are

still running and I think once we have confirmation from layer 2 people that they started um organically inducing load then we'll slowly reduce our spamming got it thank you any other also last week yeah last week uh Marson from the ne M team began very heavy blob testing and uh we saw an increase in memory usage in pretty much all the clients I just linked to uh the dashboard for this and we saw an saw this very high memory usage even after the blob spamming was finished so

this we we don't quite know what the reason for this was probably just the block pool was absolutely hammered for all this period of time but uh yeah we had a bunch of clients just all and crashing because uh we have 8 gigs of RAM for most of these machines and they were just not enough for uh for this small for 12 so this if if you're going to see some spamming on the mainat uh we might going to see some big problems where uh those machines that run with 16 gigs of RAM uh will not be able to keep up and they're

just going to ow and but that's also going to be a very expensive uh operation on Main net because yeah keeping six blobs every slot is going to be not cheap right it's exponentially more expensive right exactly yeah yeah and I think there was something specific that the red team had to look into as well right yeah but they push out the fix and okay cool like we had we had issues uh with gas we had issues with uh with Aron or noton sorry uh with yeah R you said and uh with bz also so it was it was mostly the execution layer clients but

also just see clients also ramped up in R usage During the period yeah and one other thing um we noticed was that nethermind wasn't respecting the M Max blobs to store um so they were just basically eating up dis space until they just didn't have any and they fixed that now now they were using like 50 gigs more than every other client yeah and it was a clear regression so should be fine now I think it said nether

mind yeah yeah I think M posted about it on interrup I'll just try to find them message yeah yeah we mainly nost because yeah uh another thing is We Begin spamming not spamming sorry uh we started depositing 25,000 V lits into G so this is just to check uh the main default for the max per AO activation turn limit which should be set to8 So currently it is 8 on girlyy and it will be nine with uh 500 89,000 or something

like that validators and we going to try to reach that in the oncoming weeks and Bas on our calculations it should happen right after sapoia and right before hsky uh Forks uh than yeah the deposits are already done um if if you check a pending que do like 24,000 valid do pending right but we wouldn't reach if we see some Mass exits so we would like ask every team not to exit any of their validators until we can test this uh turn limit so that's like another two weeks or

so um and do we know the when the first blob will expire uh it should be the 8th of February okay nice um so right after so right after our last no right after yeah right after okay right after our first test net goes live right okay um was it like fifth of I thought it was fifth no actually yeah right I think it's fifth even um yeah it's fifth yeah I think I posted a message somewh actually no sorry so this will be right after hosi not so uh sepolia is supposed to Fork on the 30th right so it'll be right after or

potentially between our two test Nets and potentially after the last one yeah um I think we basically had everything in play to make a decision on the 8th of February about me andet okay so the 5th of February is going to be probably expired and the 30 we fork zolia and the week after that we Fork uh if I remember yeah so we should have expired blobs by the time we Fork hki yeah exactly we should have the deposits processed by the time we also for

Hy so we should be good with Hy Hy should just a c to check at that point sweet um okay and Alex is saying there seems to be some issue with blob scan um yeah it's on our side the issue is purely on our side they have a g instance running somewhere yeah just looking last night and some things looked off or just were broken but as long as we're aware that's like the main thing we have I think to

look at Bobs without hitting API so it yeah blob scan.com just resolves not gly yes so if if you go to blob scan.com you will get uh in top left corner you can see that it's SS to or the network is aced early it was I think that stat stuff is broken still they're not showing any of the statistics yeah they said that they are fixing that today it was some reorg issue that they were not able to process and they have some chart issue also it should be all fixed uh by today hopefully and our RPC end point is offline now for DEET

12 because I had to resync uh the RPC endpoint and I yeah it hasn't finished yet sweet um anything else on testing and general Readiness that people want to cover are all the teams going to have a release out uh today or tomorrow

ideally okay I'll take the silence as a yes um and yeah are we expecting one release now for both Forks correct one release for both test Nets um yeah and we both uh yeah we anounce both in the same blog post can we then just the merge in the hosy PRS which hoski PRS yeah I think there's just this one outstanding oh yeah there's just one everything else should have already been no no everything else has been

marched in um this one's just a modification to set the um Max EO CH limit to 8 which is the same as main net got it it's yeah we just didn't have it present yeah okay the field was missing earlier yeah let's definitely merge it uh yeah it's been open for four days um yeah I say let's merge it today or at the latest tomorrow morning if there's unless there's any objection on it um yeah who can merge this who's like a I should have okay it's merch

awesome oh okay uh Mario has an update about invalid block test invalid block okay not to be testing uh on clients sweet and then yeah I would definitely keep our call two weeks from now so we can review how the test Nets went uh potentially have some blobs expire depending on the exact time that happens um and then we'll see after that one if we want another one two weeks after that that might be right before main net so there might be value or um yeah let's let's figure it out then I think there's

one other dog if we have the time I just want to go through um which is the attack net talk so this is the tool where we can control a lot more parameters in the network so you can imagine you can do Network latency packet drop split Network views um messing with time on the network and stuff like that um there's a couple of test scenarios we wrote down that we think would be good use but um yeah the question is still what's most useful for the Cent teams and what data do you guys want us to collect and how do you want us to show it to you and so on so it would be great if you can have a look at the talk and then um either DM me or leave comments or whatever you

can sweet any questions comments okay and yeah it's probably worth posting a tacet more broadly in the R&D Discord somewhere um so there's more ice on it um we'll do okay anything else before we wrap up when do we start testing Spectra soon as we have uh

implementations we already have three eips that have been CFI so you know I'm just waiting on client teams any news from client teams on that uh no I doubt anyone would seriously have started I was joking to be clear have we heard back from any more l2s about when they're using 4844 on go/ other networks cuz I think we're also starting to see some more messages about hey my library isn't working and so on so I think starting earlier makes more sense

that is the op DEET going to be on girly or yeah girly because it's the only one that's forked right now right sweet anything else uh before we wrap

up okay well well thanks everyone um I'll be monitoring all the releases pages on GitHub and pinging you tomorrow uh midday if I haven't seen anything there yet um and yeah let's get the last few test Nets done talk to you all soon see you thanks bye everyone byebye