add non-durable atomic write option#9
Conversation
Signed-off-by: sallyom <somalley@redhat.com>
|
Codex review: needs real behavior proof before merge. Summary Reproducibility: not applicable. as a fs-safe bug reproduction; this is a new public API option. Source inspection does verify that current main has no Real behavior proof Next step before merge Security Review findings
Review detailsBest possible solution: Land a narrow fs-safe API update that keeps durability enabled by default, exposes the opt-out only where maintainers want that contract, documents the tradeoff, and gives OpenClaw a released version to consume. Do we have a high-confidence way to reproduce the issue? Not applicable as a fs-safe bug reproduction; this is a new public API option. Source inspection does verify that current main has no Is this the best way to solve the issue? Mostly yes: defaulting Full review comments:
Overall correctness: patch is correct Acceptance criteria:
What I checked:
Likely related people:
Remaining risk / open question:
Codex review notes: model gpt-5.5, reasoning high; reviewed against 55327c893031. |
Supports the OpenClaw fix for gateway stalls reported in openclaw/openclaw#73655, where frequent session-store writes held the session-write lock across fsync and caused long lock holds, event-loop starvation, polling stalls, cron timeouts, and slow session/status commands. This adds an opt-out for expensive fsync work on atomic text/JSON writes so hot, reconstructible metadata can keep atomic replace behavior without blocking latency-sensitive paths.
openclaw PR from before fs-safe: openclaw/openclaw#76881