Issue #1157: drive.nika: a restart of the container kills the driver and its builder without a trace

Open and answer this issue on GitHub.

ready · opened 2026-11-02 · updated 2026-11-02 · 0 comments

Found during the night of 2026-11-02: around 05:48 the cloud container restarted (uptime was 48 s at 06:41). The driver and the S2 builder were gone, and nothing in the log said so. About 1 h was lost until the next check-in.

Resuming already works (docs/briefs/<brief>.session, the “interrupted” prompt). What is missing is noticing the restart.

To build

1. The driver writes a heartbeat (time, brief, phase) to a file next to the log.
2. drive.nika --status (or the check-in) reads it and reports “driver dead since X” when the process is gone and the heartbeat is old.
3. Document in docs/supervisor.md that a check-in starts the driver again when it finds it dead.

Back to Open Issues