Expand description
The last stop, kept where it survives the stop (2026-10-01, the real test2 run 9:
vm1 stopped cleanly, but its power-off steps were lost — the narration and the disk log
die with the read-only remount of /data — and UpCloud reported stopped 46 s after the
soft stop, so nobody could tell the guest’s power-off from the provider’s reporting).
After the quiesce and just before the power goes, ONE line is written on the ESP (its
own filesystem, flipped read-write for that write); the next boot says it at INFO and
removes it.
Constants§
- RECORD
- Where the record lives on the ESP.
- SAID
- The words the next boot says the record in — what the installer’s kit looks for.
Functions§
- keep
- Write the record for a stop said as
line, stamped now — the last write before the power. - power_
off_ epoch_ ms - The power-off time out of a record (or out of the line the next boot said it in).
- render
- The record:
power_off_epoch_ms=<ms> <the stop's line>— one line, no newline in it. - say_
previous - At boot: say the previous stop’s record (if any) at INFO, then remove it.