fix(P09): migration + operational safety — job stop, retention, logs cap (REQ-158)
- job stop: real systemctl stop via SSH (was DB-only soft stop)
resolves node from alloc_history or --peer flag
- doctor db-retention: row count check for jobs/tasks/audit_log
warns at 100k rows, suggests backup + cleanup
- logs --lines: cap at 50000 (default 1000); --since upper bound 7d
prevents OOM from unbounded journalctl
- cache DB mode 0600 (was 0644; matches store.Open)
- upgrade cutover: backup file + atomic rename (was sed -i)
rollback restores from backup on failure
Tests: job stop SSH, DB retention warning, logs lines cap, cache mode,
cutover backup-restore + atomic rename.
---ci---
project: orca
phase: 9
milestone: v0.13
status: complete
requirements:
covered: [158]
---/ci---
Jon Chery
531b36924c