fix: increase block production timeouts to account for event loop lag #7420
Conversation
This file contains 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
nflaig
commented
Feb 2, 2025
| */ | ||
| const BLOCK_PRODUCTION_RACE_CUTOFF_MS = 2_000; | ||
| const BLOCK_PRODUCTION_RACE_CUTOFF_MS = 2_500; |
There was a problem hiding this comment.
In my opinion increasing this by 500ms should be fine because the previous value was absolute, meaning no matter at which time we started the block production race it would always give 2 seconds while now it's most of the time just 1.5-1.7 seconds since the cutoff it based on seconds into slot.
We also have a strict timeout on builder header request now which means this cutoff should rarely be relevant.
In addition, this value was chosen when we used JSON to pass the block around vc<>bn which added at least 500ms more to publish flow compared to using SSZ + we no longer flood publish blocks and blobs which makes it even less likely to orphan blocks.
There was a problem hiding this comment.
pass the block around vc<>bn which added at least 500ms more to publish flow compared to using SSZ
ok this makes sense to me. so lgtm
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## unstable #7420 +/- ##
=========================================
Coverage 50.26% 50.26%
=========================================
Files 602 602
Lines 40376 40376
Branches 2206 2206
=========================================
Hits 20293 20293
Misses 20043 20043
Partials 40 40 |
Performance Report🚀🚀 Significant benchmark improvement detected
Full benchmark results
|
wemeetagain
approved these changes
Feb 3, 2025
philknows
pushed a commit
that referenced
this pull request
Feb 3, 2025
…#7420) **Motivation** Based on observations from mainnet nodes, it seems like we reject builder blocks in some cases either due to not being sent within cutoff time or due to timeout on the api call but looking at the response times these have been timely. The reason why those were rejected is either we started the block production race too late into the slot, which is mostly due to the fact that we take too much time to produce the common block body or the timeout was handled by the node with a delay, both of these cases are likely caused by event loop lag either due to GC or processing something else. See [discord](https://discord.com/channels/593655374469660673/1331991458152058991/1335576180815958088) for details. **Description** Increase block production timeouts to account for event loop lag
|
🎉 This PR is included in v1.26.0 🎉 |
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.
Motivation
Based on observations from mainnet nodes, it seems like we reject builder blocks in some cases either due to not being sent within cutoff time or due to timeout on the api call but looking at the response times these have been timely. The reason why those were rejected is either we started the block production race too late into the slot, which is mostly due to the fact that we take too much time to produce the common block body or the timeout was handled by the node with a delay, both of these cases are likely caused by event loop lag either due to GC or processing something else.
See discord for details.
Description
Increase block production timeouts to account for event loop lag