Server Health & Maintenance

“Live resource usage, service/queue/scheduler status, and confirmed restart actions for this PatchPilot instance.” This is the operational page for keeping a self-hosted instance running — found at Settings > Server Health, three tabs.

Resources

Host-level metrics: CPU / memory / disk usage over time, and whether the host has a pending reboot (kernel or library update waiting). Two independent controls live here, and the page is explicit that they are different things:

  • Docker live-restore — keeps containers running across a dockerd restart. Turning it on for the first time still causes one brief all-container restart; after that, a Docker Engine update no longer restarts every container at once.
  • Restart Server (OS reboot) — reboots the whole virtual machine, OS and kernel included, not just the containers. Everything, including the dashboard itself, is briefly unreachable (about a minute). This is deliberately separate from “Restart entire stack” on the Containers tab, which only restarts containers and never touches the OS.
  • An optional auto-reboot schedule (server time, UTC) can apply pending OS updates automatically at a set hour.

Workers & Schedulers

Read-only visibility into the background processing layer:

  • Workers — BullMQ queue depth per queue (Remediation / Schedules / Reports), worker-process liveness, and a stuck-jobs tile.
  • Schedules — the same recurring schedules configured on the Operations > Schedules page, shown here from an operational angle (last/next fire time, health) rather than a configuration one.

There is deliberately no restart button on this tab — restarting the api or worker process now goes through the same queued, audited path as the Containers tab below, instead of a separate one-off action.

Services & Containers

  • Services — status of the app’s own dependencies (Postgres, Redis, and similar).
  • Containers — a queued, audited restart for an individual container, or Restart entire stack (every container except the updater sidecar itself). Restarting a single container goes through the updater sidecar and is not instant; a stack restart briefly takes the whole instance down. Every restart, of either kind, is recorded in the Audit Log.
  • Restart actions need the updater sidecar (unavailable in Demo Mode) and settings write access — without either, the controls are hidden or disabled rather than erroring.

Backups

Postgres is dumped nightly by the backup container, on a schedule controlled by two environment variables:

Variable Meaning
BACKUP_HOUR Hour (0–23, container-local time) the nightly dump runs. Default 2.
BACKUP_RETENTION_DAYS How many days of dumps to keep before the oldest are deleted. Default 14.

Dumps land on local disk only — /opt/patchpilot/backups on the Azure VM deploy path, or the host’s own disk otherwise. This is by design and does not survive losing the machine. Arrange your own off-box copy (for example a scheduled az storage blob upload-batch via run-command on the Azure path) if that matters for your deployment.

See Requirements for the full environment-variable reference, and Azure Deploy for the Azure-specific update and troubleshooting commands.

This site uses Just the Docs, a documentation theme for Jekyll.