rehome: node-restart aftercare (uncordon, exact Ready, wait for the forge) #260

Merged

plat merged 2 commits from rehome-node-aftercare into main

2026-09-02 04:51:00 +00:00

plat commented

2026-09-02 03:36:05 +00:00

Owner

The gauntlet’s rehome scenario (run 846, issue #258) got through germinate + the full gate + the rehome itself, then died in the k3d mirror-restart aftercare: a docker-restarted node returns Ready,SchedulingDisabled — the readiness grep matched the “Ready” substring, forgejo could not reschedule onto the cordoned node, and the port-forward starved. Fix: uncordon every node after the restart, require the status column to be EXACTLY “Ready”, and wait for the forgejo rollout before dialing it. The next gauntlet run is the proof.

The gauntlet’s rehome scenario (run 846, issue #258) got through germinate + the full gate + the rehome itself, then died in the k3d mirror-restart aftercare: a docker-restarted node returns **Ready,SchedulingDisabled** — the readiness grep matched the “Ready” substring, forgejo could not reschedule onto the cordoned node, and the port-forward starved. Fix: uncordon every node after the restart, require the status column to be EXACTLY “Ready”, and wait for the forgejo rollout before dialing it. The next gauntlet run is the proof.

plat added 1 commit

2026-09-02 03:36:05 +00:00