EIPsInsight

Loading Experience

Preparing your insights...

Loading0%
PQI2026-06-10_043

PQ Interop #043

2026-06-10 2 decisions 257 transcript lines

Transcript

Ask about this call

Answered from this call's summary and transcript only.

AI-generated — verify against the recording.

Call summary

Targets
  • Lantern devnet-5 image with leanMetrics PR #36 metric — next image release (imminent) — 00:33:47
Decisions
  • devnet-4 public interop runs discontinued; devnet-5 is the active target for all multiclient testing — 00:35:42
  • 8-attestation-data entries per block is a maximum, not a requirement; proposers may include fewer to meet proof deadline — 00:30:58
Highlights
  • Client Updates:
    • ·Qlean: devnet-5 image released; Qlean-to-Qlean finalization not yet achieved — 00:14:00
    • ·Ethlambda: RocksDB descriptor-limit fix, LeanVM bumped to devnet-4 latest, invalid attestation bit rejection added; EL integration with ethrex ongoing — 00:15:00
    • ·Gean: LeanVM bump merged; devnet-5 stabilization in progress against Pata's reported issues — 00:16:14
    • ·Grandine: devnet-5 changes under local testing; image publish imminent — 00:18:03
    • ·Lantern (chat): devnet-5 image on commit 8fcd; locally updated to 3f2116d; gossip decoding error (reported by Pata) root-cause investigation underway — 00:14:29
  • Testing Progress:
    • ·devnet-4 multi-subnet interop abandoned; all teams now focused on devnet-5 — 00:10:44
    • ·Zeam: devnet-5 single-subnet finalization confirmed; 4-subnet runs in progress with late message publishing under investigation — 00:11:32
    • ·Ream: 24-node/4-aggregator/6-node-per-subnet devnet-4 infra running; memory growth resolved; finalization still bursty — 00:12:30
    • ·Pata: 32-node Zeam-only 4-subnet devnet-5 run progressing; several issues already reported to teams — 00:31:47
  • Network Performance:
    • ·64-node single-subnet sim: aggregation topic generating ~60 MB/slot per node due to GossipSub duplicates; tuning needed — 00:34:21
    • ·Block propagation P95 ~2 s at 50 Mbps bandwidth cap (EIP-7870 baseline); slot duration increase (4 s → 6 s) or interval rebalancing proposed — 00:36:04
  • Lean Spec And Leanvm:
    • ·lean spec refactored: internal py tests removed, replaced by consumable test vectors with uniform given/when/then structure — 00:18:22
    • ·CI time cut from ~40 min to ~4 min by mocking proofs/verification in non-crypto tests; real crypto tests isolated to dedicated folder — 00:23:09
    • ·Monolithic spec.py split into separate files: state transition, fork choice, etc. — 00:25:50
    • ·LeanVM: ZK allocator now available without manual global-allocator setup; work begun on making LeanVM genuinely ZK — 00:28:10
    • ·SHA-256 arithmetization for LeanVM underway (grant contributor Onyur); targeting precompile with lookup constraints — 00:27:15
Action Items
  • All lean client teams — All client teams: include aggregator-slot metric from leanMetrics PR #36 in next devnet-5 images — 00:30:29
  • All lean client teams — Test locally with 12 GB container RAM limit to match devnet server constraints — 00:32:50

Meeting chat18

  • Mihir Faujdar

    Hey everyone, my audio is not working properly currently so I would like to post an update for Lantern in the chat. For devnet-4, local testing was done for multi-subnet (two subnets) setup using lean-quickstart which helped to surface message issues in Lantern. For devnet-5, we have been keeping up with the new performance enhancements being made to the leanVM repo. The public devnet-5 image for lantern is still using the 8fcd commit. Locally i have updated it to use 3f2116d commit. Urgent priority is to find the root cause for gossip decoding error reported by Partha. Trying to replicate the issue locally before confirming the fix for it.

  • Parthasarathy Ramanujam

    Reacted to "Hey everyone, my aud..." with 👍

  • Mega | Lambda

    Reacted to "Hey everyone, my aud..." with 👍

  • Mihir Faujdar

    Sorry my audio is not working today @Thomas Coratger , have posted my update above.

  • Thomas Coratger

    No problem, thanks a lot!

  • Mihir Faujdar

    Reacted to "No problem, thanks a..." with 👍

  • Ladislaus

    Hat auf "Hey everyone, my aud..." mit 👍 reagiert

  • Parthasarathy Ramanujam

    https://github.com/leanEthereum/leanMetrics/pull/36 can teams ensure this metric will be included in the devnet5 images?

  • Mihir Faujdar

    Confirming Lantern will support this new metric in the next image for devnet-5.

  • Parthasarathy Ramanujam

    Reacted to "Confirming Lantern w..." with 👍

  • Mega | Lambda

    Will we still do devnet 4 runs in public interop?

  • Gajinder Singh

    Replying to "Will we still do dev..." no

  • Mega | Lambda

    Reacted to "no" with 👍

  • Pablo Deymo | Lambda

    Reacted to "no" with 👍

  • Gajinder Singh

    Replying to "Will we still do dev..." we have already started devnet5 multiclient interops, things seemed good and progressing well for 1 subnet for 5 clients I think. Partha can tell more about this

  • Mega | Lambda

    Replying to "Will we still do dev..." Good to know. We’ll merge devnet5 to our main branch then

  • Gajinder Singh

    Reacted to "Good to know. We’ll ..." with 👏🏾

  • Mihir Faujdar

    Thanks everyone!

Key decisions

  • devnet-4 public interop runs discontinued; devnet-5 is the active target for all multiclient testing

    devnet-5
  • 8-attestation-data entries per block is a maximum, not a requirement; proposers may include fewer to meet proof deadline

    clarifies the maximum number of attestation data entries per block