Repository navigation
Conversation
…ur L1 yet A light node checks the upstream CurrentL1 against its own L1 RPC. When that RPC is a few blocks behind the upstream's (for example a consensus-aware L1 proxyd whose head moves back while a lagging backend rejoins), the block is not found and the whole poll was rejected and counted as error_l1_lookup, discarding safe and finalized heads that were already verified. Treat a not-found CurrentL1 like an upstream that reports none: apply the verified heads, leave CurrentL1 unchanged, and count it in the new follow_source_current_l1_not_found_total metric. A CurrentL1 hash mismatch or any other lookup error still rejects the poll. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Light nodes (
--l2.follow.source) reject a whole follow-source poll, and count it aserror_l1_lookup, when their own L1 RPC doesn't have the upstream's CurrentL1 yet. On the staging devnets, the L1 endpoint is a consensus-aware proxyd whose consensus head moves back a few blocks when a lagging L1 backend rejoins. That makes this happen in bursts on every light node at once, and it fires thel2-follow-source-error-rate-*alerts. From Nov 1–8, all 1,309 follow-source errors on sepolia-devnet-2, sepolia-devnet-3 and platforms-1 were this case, and every one fell in an L1 backend lag window.After this change, the node applies the already-verified safe and finalized heads, leaves CurrentL1 unchanged, and increments
op_node_default_follow_source_current_l1_not_found_totalinstead of recording an error. A CurrentL1 hash mismatch or any other lookup error still rejects the poll. A matching CurrentL1 is what proves the L1 blocks holding the batches are on our chain, so those cases stay strict.Effect on operators: follow-source error rates drop, and there is one new counter. A follower whose L1 RPC stays behind indefinitely no longer shows up in the error rate, so an alert on this counter belongs in the alert rules. That's a follow-up.
Reviews:
go-code-reviewerran. Its main suggestion, keeping mismatches and other lookup errors strict, is adopted.safedb-reviewerdoesn't apply: follow mode promotes safe heads with an empty L1 source either way. The kona-node equivalent is #23990.🤖 Generated with Claude Code