Repository navigation
Keep resident WAL catch-up scheduled without external refill - #329
Merged
Merged
Conversation
beinan
marked this pull request as ready for review
October 8, 2026 07:22
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Resident masters can finish their 16-pass catch-up slice while WAL remains, then wait for a coarse stats scan or an external agent to enqueue another task. This change persists continuation demand before the slice returns and adds a bounded manifest-only discovery loop for already-owned local append targets, independent of full stats scans.
The discovery loop shares one fleet-wide reservation, rotates persistent cursors, uses at most two metadata readers, preserves external owners and canonical claims, and respects the durable catch-up failure deadline. It never replays raw WAL, resets ownership, changes rollout allowlists, or cancels legacy handlers. Dedicated executor mode retains its existing scheduling behavior.
Validation: all five focused local tests passed against isolated etcd, and all 11 CI checks passed, including Rust unit/integration/HA tests, coverage, Clippy, and Python tests. Tests cover replica deduplication, missing-table isolation, cursor recovery, external-owner preservation, retry backoff, and a 17-generation backlog spanning two resident slices without the scanner. See docs/resident-wal-recovery.md for operational limits; legacy cancellation and flush scheduling remain separate work.