Report training record latency against the block's sealing time - #2
Open
kimminhyun-ai wants to merge 1 commit into
Open
kimminhyun-ai wants to merge 1 commit into
kimminhyun-ai wants to merge 1 commit into
Conversation
A block's `timestamp` is when its proposer STARTED building it, and transactions keep arriving into that block for the next second or two. So a training record's reported `submitted_at` is routinely later than the timestamp of the very block that carries it. The latency was `block.timestamp - submittedAt`, which went negative, and the guard against negatives turned that into "Unavailable". On a ten-validator chain carrying seventy parallel training pipelines, 48 of 70 records landed in exactly that position: a real 0.8-to-3.7-second latency read as no data at all. Arrival is now the moment the containing block was sealed — the next block's timestamp, which is when this block stopped accepting and the next began. With no next block there is no arrival time, and the answer stays null rather than becoming a guess. The page fetches that block and both the detail row and the batched-record list explain what the number is measured between. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019hZqdLKnVfzvRf9mus6NNh
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
A block's
timestampis when its proposer started building it, and transactions keep arriving into that block for the next second or two. So a training record's reportedsubmitted_atis routinely later than the timestamp of the very block that carries it. The latency wasblock.timestamp - submittedAt, which went negative, and the guard against negatives turned that intoUnavailable.On a ten-validator chain carrying seventy parallel training pipelines, 48 of 70 records landed in exactly that position — a real 0.8-to-3.7-second latency read as no data at all, on the one screen the certification procedure points at for this metric.
Arrival is now the moment the containing block was sealed, read as the next block's timestamp: when this block stopped accepting and the next began. With no next block there is no arrival time and the answer stays
nullrather than becoming a guess. The transaction page fetches that block, and both the detail row and the batched-record list say what the number is measured between.Verified against a live 10-validator chain: a READY record that previously showed
Unavailablenow shows834 ms.🤖 Generated with Claude Code
https://claude.ai/code/session_019hZqdLKnVfzvRf9mus6NNh