fix(agent): oneshot timer units must not flap the node to degraded #93

Merged

gmackie merged 1 commit from fix/degraded-oneshot into main

2026-07-12 20:32:25 +00:00

gmackie commented

2026-07-12 20:26:09 +00:00

Owner

Observed live: hetzner-master reported degraded because fg-retention/fg-monitoring-probe (timer-driven oneshots) are inactive between runs, and any inactive fg-* unit flips the whole node. Skip Type=oneshot units.

🤖 Generated with Claude Code

gmackie added 1 commit

2026-07-12 20:26:09 +00:00

fix(agent): oneshot timer units must not flap the node to degraded

CI / ci (pull_request) Successful in 5m32s

Details

6cf5fee0fd

fg-retention / fg-monitoring-probe are timer-driven oneshots — inactive
between runs by design. The heartbeat's degraded computation counted any
inactive fg-* unit as down, so a node went 'degraded' whenever a timer
job's unit was still loaded (observed live on hetzner-master right after
the 0.1.29 rollout). Skip Type=oneshot units from service snapshots.

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

gmackie merged commit ba685633d1 into main

2026-07-12 20:32:25 +00:00

gmackie deleted branch fix/degraded-oneshot

2026-07-12 20:32:25 +00:00

gmackie referenced this pull request from a commit

2026-07-12 20:32:26 +00:00

fix(agent): oneshot timer units must not flap the node to degraded (#93)

gmackie referenced this pull request

2026-08-27 00:56:49 +00:00

feat(web): web run reports — consume @preflight/runreport #340

gmackie referenced this pull request

2026-08-27 01:07:18 +00:00