Repository navigation
Planned llama restarts opt-in; failsafe watchdog; single instance - #15
Merged
Merged
Conversation
…ama-server only Incident 2026-10-05: Hawk restarted llama-server on a fixed cadence (every 3.5 h) without any switch, and after Hawk itself died nothing was left to bring either back. - planned restarts only when llama_kill_degraded is on and the average decode speed is below llama_kill_tps; no time cadence by default (0); a dead server is always started again - hawk ensure / hawk install-watchdog: Task Scheduler (pythonw, every 2 min) or cron starts a stopped Hawk and a llama-server down for two checks - OS file locks: one monitor, one dashboard, one llama-server start at a time (the alive check is repeated under the lock) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Follow-up to the 2026-10-05 overnight outage.
llama_kill_degradedis on and the average decode speed is belowllama_kill_tps; the time cadence defaults to off (0). A dead or unresponsive server is always started again.hawk ensurestarts a stopped Hawk (cleanly, never twice) and a llama-server that stays down for two checks;hawk install-watchdogruns it every 2 minutes via Task Scheduler (pythonw, no window) or cron.Tests: planned_due, ensure failsafe, single-instance lock across processes; all suites green locally.
🤖 Generated with Claude Code