Logflare / logflare / 36458417052-1

83%

main: 83%

Build Type

Pull #4067

github

Committed by Baishan

Commit Message

fix: flush the disk rotating WAL on an idle timer, not just on commit

RotatingWal only rotated (sealed the active segment and shipped it to
the inner backend) from inside commit/4, so the time-based rotation
check never ran once traffic stopped — a pending tail would sit in the
local WAL file until the next append or a clean shutdown, never
clearing during a drain.

Track the interval on a `:timer.send_interval/2` tick delivered to the
owning process as `{:backend, :flush_tick}` and routed through the
existing `handle_message/2` hook, forcing a rotation once
max_rotation_interval_ms has elapsed since the last one, independent of
whether new appends are arriving.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

Pull Request Pull Request #4067: fix: flush the disk rotating WAL on an idle timer during drain

Coverage Stats

5 of 5 new or added lines in 1 file covered. (100.0%)

1 existing line in 1 file now uncovered.

15733 of 18894 relevant lines covered (83.27%)

2085.43 hits per line