AllCoreDevs — Testing #069
Transcript
Ask about this call
Answered from this call's summary and transcript only.
AI-generated — verify against the recording.
Call summary
Targets
- •Mid-week - ETH/70 isolated testnet launch — 00:49:05
- •End of March - Final gas repricing numbers after BAL benchmarks — 00:58:23
Decisions
- •Reduce bal-devnet-2 gas limit to 70M to focus on BAL testing — 00:31:21
- •EPBS DevNet 0 will not test fork choice edge cases — 00:38:31
- •ETH/70 should be tested in isolation before adding to BAL devnets — 00:47:32
- •EIP-8037 (state growth) prioritized for DevNet 3 implementation — 00:59:08
Highlights
- Bal Devnet 2 Status:
- ·bal-devnet-2 launched Friday; Nethermind and Nimbus EL have syncing issues — 00:17:58
- ·Nimbus issue relates to block access list for finalized blocks — 00:19:33
- ·Network configured at 150M gas; clients should verify compatibility — 00:21:47
- ·10MB block limit active (Fusaka); constrains blocks above ~104M gas — 00:22:03
- Eth70 And Gas Limits:
- ·ETH/70 implementation ready in Besu; testing against other clients needed — 00:43:51
- ·ETH/70 kurtosis testing shows clients still syncing unexpectedly — 00:45:25
- ·Full-sync bypasses receipts; snap-sync needed to trigger ETH/70 path — 00:46:32
- ·ETH/70 is critical blocker for scaling beyond 83M gas — 00:47:09
- Gas Repricing Updates:
- ·EIP-7904 (compute repricing) has preliminary numbers; awaiting BAL-optimized benchmarks — 00:53:42
- ·EIP-8037 (state growth) specs progressing; multidimensional metering details being updated — 00:55:24
- ·EIP-7976/7981 (data repricing) most advanced; specs complete — 00:56:11
- ·Final repricing numbers expected end of March after BAL benchmarks complete — 00:58:08
- ·EIP-8037 recommended for DevNet 3 as most complex implementation — 00:59:15
- Epbs Devnet 0 Progress:
- ·Prysm: fork choice ready in 1-2 days; API and sync remain — 00:35:12
- ·Teku: passing Alpha 2 spec tests except fork choice — 00:36:09
- ·Nimbus: Alpha 2 PR merging imminently; fork choice deprioritized for DevNet 0 — 00:37:04
- ·DevNet 0 will not focus on fork choice edge cases — 00:38:31
- ·EL testing image available: ethpandaops/geth:epbs-devnet-0 — 00:40:04
- ·Lodestar: passing Alpha 2 tests; fork choice refactoring in progress — 00:41:11
- Benchmarking Infrastructure:
- ·Benchmarkor tool released: consumes EEST fixtures and raw JSON tests — 00:50:32
- ·Benchmarkor supports stateful tests with 1.2-1.6TB snapshots via CoW filesystems — 01:02:00
- ·Benchmarkor includes UI for test heatmaps, resource graphs, and detailed logs — 01:05:36
- ·Carlos's standardized block metrics JSON format enables cross-client analysis — 01:09:20
Action Items
- •PK/Barnabas — Set up isolated ETH/70 testnet for kurtosis validation — 00:49:00
- •All EL teams — Clients reach readiness for ETH/70 testing and notify Barnabas/PK — 00:50:00
- •Maria/Stefan — Coordinate DevNet 3 scope: prioritize EIP-8037, consider data repricings — 01:00:30
Meeting chat163
- spencer-tb
Zoom bot? 👀
- Mario Vega
Reacted to "Zoom bot? 👀" with 👀
- Stefan Starflinger
my audio
- Stefan Starflinger
does not work
- Stefan Starflinger
sec
- Barnabas
micdao
- Dragan Rakita
Reacted to "micdao" with ➕
- Mario Vega
Reacted to "micdao" with ➕
- spencer-tb
Reacted to "micdao" with ➕
- Justin Traglia
Reacted to "micdao" with ➕
- wolovim
Stefan’s spec doc: https://notes.ethereum.org/@ethpandaops/bal-devnet-2
- Jared Wasinger
It should be present for non-finalized blocks though
- Barnabas
a single malicous actor could trigger a chain to halt over 100M gas without eth/70
- Dustin
10MB still matters on CL gossip
- Dustin
yes 10MB is a hard limit right now
- pk910
receipts are more problematic than block payloads >10M. a bigger block can't be broadcasted, but everyone generates the receipts themselves - except whenhitting sync
- Ben Adams
Call data price change is still important; because the size cap determining blocks rather than pricing distorts the gas market
- Marius van der Wijden
What was the point about the ef making noises again? Sorry I missed the sentence
- Toni Wahrstätter
It should be tested, yes, but we wouldn't go to 150M gas without the calldata pricing adjustment. After the calldata increase we have (a lot of) leeway again before getting close to 10mib.
- Potuz
Want to call attention that fine control on the max size is also actually needed for ePBS because it will determine the max broadcast time allocated.
- Ansgar Dietrichs
Replying to "What was the point a..." case in point :-) just kidding. he was talking about the suggestion that there might be a deadline for Glamsterdam, where even headliners that miss the deadline might be out
- Mario Vega
Reacted to "It should be tested,..." with 👍🏼
- felipe
Reacted to "It should be teste..." with 👍🏼
- Dragan Rakita
Reacted to "It should be tested,..." with 👍🏼
- Toni Wahrstätter
Replying to "Want to call attenti..." Bc of the dynamic ptc deadline?
- Stefan Starflinger
Reacted to "It should be tested,..." with 👍🏼
- Marius van der Wijden
Replying to "What was the point a..." Ah yes, I think thats a good thing, no?
- Francesco
I would agree, it’s not clear why we should have this extra cap
- Potuz
Replying to "Want to call attenti..." yes, that seems to be a needed feature in this fork given the gas numbers that are being talked about
- Caleb
Reacted to "Want to call atten..." with 👍
- Ameziane Hamlat
Reacted to "It should be tested,..." with 👍🏼
- Stefan Starflinger
Replying to "It should be tested,..." so what would you recommend for devnet-2 reduce to 100M?
- Dustin
Replying to "What was the point..." Not suggesting otherwise but saying it impacts BAL planning
- Dustin
Replying to "What was the point..." and means the existing overhead from Fusaka has to remain even though ePBS would remove most of it
- Marius van der Wijden
Replying to "What was the point a..." Yep makes sense
- Marius van der Wijden
Reacted to It should be tested,... with "👍🏼"
- Toni Wahrstätter
Replying to "It should be tested,..." 150M should be fine sice we have the rlp limit cap
- Dustin
Replying to "I would agree, it..." ok, but that's at the earliest an H* point
- Toni Wahrstätter
Replying to "It should be tested,..." Maybe not. Not sure anymore if 10mib is actually a problem ot not
- Ben Adams
Replying to "It should be tested,..." eth/70 is the missing part on going high (as mentioned receipts) https://eips.ethereum.org/EIPS/eip-7975
- Francesco
Replying to "I would agree, it’s ..." Sure, just saying I think that is the right way to do it eventually
- Barnabas
Could we focus on testing bal instead of worrying about this high gaslimit scenario?
- Dragan Rakita
Reacted to "Could we focus on te..." with ➕
- Mario Vega
Reacted to "Could we focus on te..." with 👍🏼
- Mario Vega
Reacted to "Could we focus on te..." with ➕
- Rafael Matias (skylenet)
Reacted to "Could we focus on ..." with ➕
- Toni Wahrstätter
Because there's another 10mib there?
- felipe
Reacted to "Could we focus on ..." with ➕
- Ben Adams
Reacted to "Because there's anot..." with 👍
- Barnabas
lest just decrease the gas cap to 70M and that should be safe enough to continue
- Dragan Rakita
True, put it to 100M and go with it
- Dragan Rakita
Reacted to "lest just decrease t..." with ➕
- Mario Vega
Reacted to "lest just decrease t..." with ➕
- felipe
Reacted to "lest just decrease..." with ➕
- Stefan Starflinger
Replying to "Could we focus on te..." This is about testing BALs
- Ben Adams
Replying to "It should be tested,..." Yeah, and is to allow receipt pagination
- Marius van der Wijden
Reacted to Could we focus on te... with "➕"
- Dragan Rakita
Replying to "Could we focus on te..." So we would need calldata repricing for it
- Toni Wahrstätter
Reacted to So we would need cal... with "👍"
- Marius van der Wijden
Geth miner limits blocks to ~7mb, transactions only
- Justin Traglia
@Francesco any updates on the variable PTC deadline PR?
- Ben Adams
eth/70 I think we want to have live for Glamsterdam? So would test higher than 100M at somepoint to confirm that EIP is in and works
- Marius van der Wijden
One thing I wanted to add to the prev discussion: please speed up implementation of eth/70 so we can go beyond 70mgas
- Dustin
what's max gas which fits in 10MB?
- Dustin
with repricings
- Marius van der Wijden
Replying to "eth/70 I think we wa..." We are working on kurtosis tests to show it works
- Marius van der Wijden
Replying to "eth/70 I think we wa..." Cc @pk910
- James He
We should get the beacon api prs merged in early this week
- James He
And update it with subsequent prs if needed, the discussions are too long and getting hidden
- Ben Adams
Oh is the safe limit 70 without it?
- Ben Adams
Replying to "One thing I wanted t..." Thought it was > 100M
- nflaig
Replying to "We should get the be..." working on it with Shane
- Potuz
For some reasons I can't raise hands so I'll leave it here the forkchoice change that is **needed* is to not be broken by empty slots
- Justin Traglia
Reacted to "For some reasons I c..." with 👍
- Ameziane Hamlat
Replying to "One thing I wanted t..." I see 83 mgas on EIP-7975: eth/70
- nflaig
Replying to "We should get the be..." we will remove block production changes so we can discuss that separately and move ahead with the PR and merge it
- Barnabas
we have a ethpandaops/geth:epbs-devnet-0
- James He
Reacted to "we will remove block..." with 👍
- Caleb
Reacted to "we have a ethpand..." with 👍
- Justin Traglia
Reacted to "we have a ethpandao..." with 👍
- Mario Vega
Reacted to "we have a ethpandao..." with 👍
- Marius van der Wijden
Replying to "One thing I wanted t..." Yeah 83
- Dustin
Reacted to "we have a ethpand..." with 👍
- Dustin
Replying to "we have a ethpand..." is this from some particular gh repo/branch?
- Dustin
Replying to "we have a ethpand..." or it is the gh org it looks like
- Barnabas
Replying to "we have a ethpandao..." its a fork of mine
- Barnabas
Replying to "we have a ethpandao..." I’ll check with marius if they can do an official upstream branch
- Dustin
Replying to "we have a ethpand..." that would be useful, not to rely on only a docker image
- Toni Wahrstätter
Replying to "Want to call attenti..." My current feeling is that it should just put as late into the slot as possible, ensuring max time for propagation while still ensuring our target gas limit can be executed (in the worst case) until the next slot.
- Barnabas
@FLCL mind giving us an update?
- FLCL
sorry what was the question?
- Barnabas
eth70
- Mario Vega
eth70
- FLCL
mic issues, no significant updates
- Mario Vega
Reacted to "mic issues, no signi..." with 👍🏼
- FLCL
We tried to test it and seems better to check it in a devnet
- Barnabas
“Its working” “We are not sure why its working tho” 😂 oops
- Mario Vega
Reacted to "“Its working” “We ar..." with 😂
- Dragan Rakita
Reacted to "“Its working” “We ar..." with 😂
- Barnabas
Could we just do eth70 for devnet 3?
- pk910
Reacted to "“Its working” “We ar..." with 😂
- Mario Vega
Reacted to "Could we just do eth..." with 👀
- Ben Adams
Reacted to "Could we just do eth..." with 👍
- FLCL
Btw do you guys want to split only big receipts? Or anything that can be placed under size limit?
- nflaig
Reacted to "we have a ethpandao..." with 👍
- FLCL
Replying to "Could we just do eth..." Others will be able to sync with full sync only
- Louis
Benchmark Release: https://github.com/ethereum/execution-spec-tests/releases/tag/benchmark%40v0.0.7 Benchmark Documentation: https://github.com/ethereum/execution-specs/blob/forks/amsterdam/docs/writing_tests/benchmarks.md
- Marius van der Wijden
Replying to "Btw do you guys want..." We split only when a packet would exceed the size
- Maria Silva
https://github.com/misilva73/evm-gas-repricings/blob/main/reports/gas_repricings_todo.md
- wolovim
- Louis
Replying to "https://github.com/m..." I will complete the eip-2780 benchmark today. wip
- Louis
Guru is working on eip2780 spec right now, I am working on benchmark for it
- Mario Vega
Reacted to "Guru is working on e..." with 👍🏼
- spencer-tb
Reacted to "Guru is working on e..." with ❤️
- felipe
Reacted to "Guru is working on..." with 👍🏼
- Marius van der Wijden
+1 on 8037 on devnet-3
- Dragan Rakita
Replying to "+1 on 8037 on devnet..." I like this too
- Marius van der Wijden
Replying to "+1 on 8037 on devnet..." Its the most complex one, so we should get it out of the way early
- Stefan Starflinger
Replying to "+1 on 8037 on devnet..." me too we should discuss what we want to see in devnet-3 https://forkcast.org/devnets/
- Mario Vega
Reacted to "+1 on 8037 on devnet..." with 👀
- spencer-tb
Replying to "+1 on 8037 on devnet..." Will push in EELS, tests etc this week
- Mario Vega
Reacted to "Will push in EELS, t..." with 🔥
- Stefan Starflinger
Replying to "+1 on 8037 on devnet..." 8037 is pretty large so maybe just that?
- Maria Silva
Reacted to "Will push in EELS, t..." with 🔥
- Dragan Rakita
Replying to "+1 on 8037 on devnet..." Who is the person that I can contact related to it?
- Mario Vega
Great logo 🔥
- Kamil Chodoła
Reacted to "Great logo 🔥" with 🔥
- Maria Silva
Replying to "+1 on 8037 on devnet..." You can message me Dragan
- Dragan Rakita
Reacted to "Great logo 🔥" with 🔥
- Dragan Rakita
Replying to "+1 on 8037 on devnet..." Cool, will do that
- spencer-tb
Reacted to "Great logo 🔥" with 🔥
- Stefan Starflinger
Reacted to "Great logo 🔥" with 🔥
- Toni Wahrstätter
8037 + the data repricings (7976, 7981), which have specs and aren't complex based on steel's assesment
- Louis
Reacted to "Great logo 🔥" with 🔥
- Toni Wahrstätter
Would be great to get to testing higher gas limits without running into limits
- Toni Wahrstätter
Reacted to Great logo 🔥 with "🔥"
- Maria Silva
Reacted to "8037 + the data repr..." with 💯
- nixo
Reacted to "Great logo 🔥" with 🔥
- Toni Wahrstätter
Reacted to +1 on 8037 on devnet... with "👍"
- Justin Traglia
Reacted to "Great logo 🔥" with 🔥
- Ameziane Hamlat
This is great for EL performance benchmarking. Are you getting the results on all the runs or there is a way to have runs for warmup, and then use other runs for getting the metrics ?
- CPerezz
Awesome work @Rafael Matias (skylenet) !!! 🎉
- spencer-tb
Reacted to "Awesome work @Rafael..." with 👏
- Mario Vega
Reacted to "Awesome work @Rafael..." with 👏
- Louis
Amazing!!!
- Ameziane Hamlat
Reacted to "Awesome work @Rafael..." with 💯
- Ameziane Hamlat
Reacted to "Awesome work @Rafael..." with 👏
- Kamil Chodoła
Reacted to "Awesome work @Rafa..." with 💯
- Kamil Chodoła
Reacted to "Awesome work @Rafa..." with 👏
- Maria Silva
Reacted to "Awesome work @Rafael..." with 👏
- Caleb
beauty
- Justin Traglia
Reacted to "Awesome work @Rafael..." with 👏
- Dragan Rakita
Reacted to "Awesome work @Rafael..." with 👏
- Maria Silva
And shoutout to @CPerezz ’s work on standardizing metrics. Super useful for folks wanted to analyze state operations
- nflaig
Reacted to "Awesome work @Rafael..." with 👏
- CPerezz
Reacted to "And shoutout to @CPe..." with ❤️
- wolovim
Reacted to "And shoutout to @CPe..." with ❤️
- Toni Wahrstätter
Reacted to And shoutout to @CPe... with "❤️"
- Toni Wahrstätter
Reacted to Awesome work @Rafael... with "👏"
- felipe
Reacted to "Awesome work @Rafa..." with 💯
- felipe
Reacted to "Great logo 🔥" with 🔥
- Louis
Reacted to "And shoutout to @CPe..." with ❤️
- Luis Pinto | Besu
Reacted to "Awesome work @Rafael..." with 👏
Key decisions
Reduce bal-devnet-2 gas limit to 70M to focus on BAL testing
to focus on BAL testingEPBS DevNet 0 will not test fork choice edge cases
fork choice edge casesETH/70 should be tested in isolation before adding to BAL devnets
before adding to BAL devnetsEIP-8037 (state growth) prioritized for DevNet 3 implementation
Includedfor DevNet 3 implementation