Log rotation + retention¶

logrotate¶

Config in repo: infra/logrotate/forge.conf
Installed: /etc/logrotate.d/forge (owner root:root, 644). System logrotate runs daily, picks it up automatically.
Reinstall after edit: sudo cp /home/justinwieb/forge/infra/logrotate/forge.conf /etc/logrotate.d/forge
Dry-run / debug: sudo logrotate -d /etc/logrotate.d/forge
Force rotate now: sudo logrotate -f /etc/logrotate.d/forge
Buckets:
Monitor logs (monitor-*.log, cron-errors.log): weekly, 8 generations, 10 MB size, copytruncate
Service logs (dispatcher, home-base, inbox-webhook, telegram-*, search-index): weekly, 4 generations, 10 MB size
logs/integrations/*.log + logs/integrations/telegram/*.log: weekly, 4 generations
logs/workers/*.log: monthly, 2 generations
All buckets use copytruncate so the tailing systemd services / cron jobs don't need restart.

Script: scripts/cleanup/retention.sh
Cron: nightly at 04:15 (# forge-retention block in crontab)
What it does: moves files older than RETENTION_DAYS (default 30) into <dir>/_archive/YYYY-MM/ for these four locations:
comms/results/
comms/inbox/
tasks/completed/
tasks/failed/
logs/workers/
Log: logs/retention.log (one summary per run, === retention sweep start/done ===)
Override window: RETENTION_DAYS=14 ./scripts/cleanup/retention.sh for a one-off
Never deletes — only moves into _archive/. Justin or future cleanup pass decides cold storage / drop.