Skip to main content

Operations

Lease store​

By default leases live in a crash-safe memfile — an append-only file with atomic compaction that survives an abrupt kill. For redundancy and shared state, switch the store to PostgreSQL ([lease] store = "postgres"); Lighthouse auto-creates its tables and, once Postgres is configured, refuses to start if the database is unreachable rather than silently running stateless.

High availability​

Lighthouse HA is shared-database active-active:

  • Run two or more lighthoused engines on separate hosts, each pointing [lease] store = "postgres" at the same database.
  • Both engines answer every request. The database arbitrates each allocation with an atomic claim, so no address is ever handed out twice — pools are not split between servers.
  • Each engine keeps its own server_id; clients naturally deduplicate by server-id, so there's no virtual IP to manage. If one engine goes down, the survivor keeps serving.
  • On a routed network, point the relay at both engines (two ip helper-address entries). On a flat L2 segment, both simply listen on the subnet.

Engines heartbeat their peer status, visible via the API and the System page.

caution

Active-active removes the DHCP-server single point of failure, but the shared database becomes the new one. Run a redundant/replicated PostgreSQL (e.g. Patroni or a managed HA Postgres) for true resilience.

Backup​

There's no dedicated backup tool — back up the files directly:

  • the lease store (leases.db) and reservations.json,
  • the management database (mgmt.db),
  • the config (lighthouse.toml).

When running on PostgreSQL, rely on your database's backups for lease and reservation state.

Logging & monitoring​

The engine logs structured lines to stderr and, when [logging] file is set, to a file (required on Windows). The management API surfaces recent lines and a durable lease-event audit history, and the metrics endpoint feeds Prometheus.

Recovery & reset​

  • Admin password reset (host-local only): lighthouse-mgmtd --reset-admin --email <email>.
  • Factory reset (System → Danger zone) clears leases, reservations, audit history, device notes, and all users. Use with care.