fix(backup-drill): log restic errors in the hot leg and create the work dir 0700 #524

Sammanfogat

supernaut sammanfogade 1 incheckning från fix/158-drill-hot-restore in i main

2026-10-02 18:05:03 +00:00

supernaut kommenterad

2026-10-02 17:43:02 +00:00

Ägare

What

The first production drill (2026-10-02 17:19 UTC) passed the daily leg but failed the hot leg at restic restore (exit 1, under a second). restic's stderr was not captured, so the cause is not visible. A manual restore of the same snapshot outside the drill works (201.6 MiB, 7 s).

This PR makes the next run tell us why:

The snapshots step, the restore, the cleanup and the tar stream capture stderr and log it with the exit code on any non-zero exit. A restore failure also logs the work dir mode and uid.
The work dir is created 0700 right before the restore (the umask made it 0755).
On success the restore logs restic's summary line.

Ruled out by reading the code: a missing work dir (it is created before podman run -v). Still open: something that differs under the systemd unit.

Verification

--syntax-check, ansible-lint 0, rendered orchestrator passes py_compile. Local test with a fake podman that exits 1: the error and mode are logged and the dir is 0700.

Apply

Tag backup-drill: only the orchestrator script changes. Then run the drill by hand and read the logged restic error.

Refs #158

## What The first production drill (2026-10-02 17:19 UTC) passed the daily leg but failed the hot leg at `restic restore` (exit 1, under a second). restic's stderr was not captured, so the cause is not visible. A manual restore of the same snapshot outside the drill works (201.6 MiB, 7 s). This PR makes the next run tell us why: - The snapshots step, the restore, the cleanup and the tar stream capture stderr and log it with the exit code on any non-zero exit. A restore failure also logs the work dir mode and uid. - The work dir is created `0700` right before the restore (the umask made it 0755). - On success the restore logs restic's summary line. Ruled out by reading the code: a missing work dir (it is created before `podman run -v`). Still open: something that differs under the systemd unit. ## Verification `--syntax-check`, ansible-lint 0, rendered orchestrator passes `py_compile`. Local test with a fake `podman` that exits 1: the error and mode are logged and the dir is 0700. ## Apply Tag `backup-drill`: only the orchestrator script changes. Then run the drill by hand and read the logged restic error. Refs #158

supernaut lade till 1 incheckning

2026-10-02 17:43:02 +00:00

fix(backup-drill): log restic errors in the hot leg and create the work dir 0700

ci / ci (pull_request) Successful in 1m58s

Detaljer

497102de3c

The hot restore and the snapshot listing now capture stderr and log it
with the exit code on failure, plus the work dir mode and owner. The
cleanup and tar stream log theirs too. The work dir is created with an
explicit 0700 right before the restore.

Refs #158

supernaut tvångsskickade fix/158-drill-hot-restore från 497102de3c

ci / ci (pull_request) Successful in 1m58s

Detaljer

till 2b82a2bfe2

ci / ci (pull_request) Successful in 1m54s

Detaljer

2026-10-02 17:59:45 +00:00

Jämför

supernaut schemalade den här ändringsförfrågan för automatisk sammanfogning när alla kontroller lyckas

2026-10-02 18:00:26 +00:00

supernaut sammanfogade incheckning 232dec9fd5 till main

2026-10-02 18:05:03 +00:00