Record worker pid, host, claim time, and heartbeat metadata on running
backup jobs so operators can see which worker owns a run.
Refresh the heartbeat while rsync is active and reconcile stale running
runs when the worker heartbeat stops. Add a worker option to tune or
disable stale-run reconciliation.
Refs #11
Add a warning status for BackupRun records so successful snapshots are not
reported as failed when post-run SQL retention fails.
Keep the prune error in the run result, link the successful snapshot, and
let the management command complete with a warning instead of raising a
backup failure.
Include warning runs in backup trend summaries and add a regression test
for successful backups with failed retention cleanup.
Expose a verbose rsync output option in the Django manual backup form and
store the selected value with the queued run request.
Propagate the option through the worker, direct management command, and
rsync command builder so real backups can emit itemized changes, file-list
progress, and stats when requested. Dry-runs continue to use verbose output
by default and report that consistently in requested options.
Cover the queue, worker, view, and rsync command behavior with focused
tests.
Classify rsync failures in run results so transport issues such as exit
255 and broken pipes show clearer diagnostic hints.
Teach the worker to reconcile running dry-runs when their log already
contains a terminal rsync error, and to fail stale dry-runs after their
timeout window. This prevents failed rsync processes from leaving runs
stuck in the running state indefinitely.
Add a cancel action for queued and running backup runs. Queued runs are
cancelled immediately, while running runs are marked for cancellation and
the worker terminates the active rsync process group.
Make dry-run log paths run-specific and add a defensive default dry-run
timeout so stuck dry-runs do not remain running indefinitely.
Remove rsync exit codes from run overview tables while keeping detailed
diagnostics available on the run detail payload.
Move backup execution out of the management command into a reusable
backup runner service that can execute an existing BackupRun record.
Add queue primitives and a run_pobsync_worker command so manual backup
requests can be recorded as queued SQL state and processed outside the
web request path.
Add a worker Docker service and pobsync worker CLI alias, with tests for
queued run creation, worker execution, manual run typing, and command
mapping.
