Postgres (primary)
No signal critical tier-0 DatabaseSystem of record for control plane state, task history and workspace metadata.
- Last heartbeat
- never
- Signal source
- probe
- Host
- Home1
- Target RTO
- 25m
Why it is no signal
- self No heartbeat has ever been received from this entity.
Failover
mode: manual-
Streaming replica on VIN. Promotion is one-way until a fresh base backup is taken.
Capacity
Backup
- Schedule
- every 6h + WAL archiving
- Destination
- nas-backups
- Target RPO
- 15m
- Last success
- never
- Verification
- monthly restore drill
Availability class
Fleet-fatal · target 99.5%System of record for control-plane state, task history and workspace metadata.
- Class
- tier-0 Fleet-fatal: loss stops the estate or risks data. Everything else waits.
- RTO (declared)
- 25m
- RTO (achievable)
- 25m
- RPO
- 15m
- State
- durable
- Redundancy
- active / passive
- Failover
- manual → postgres-replica
- Failure domain
- fd-home-rack
Health check: TCP 5432 probe every 60s, plus connection, disk and replication-lag metrics — proves the listener accepts connections.
Observed independently of the service.
Blind spot: Liveness only. It does not prove the primary still accepts writes, and a full connection pool looks identical to an outage.
Without it: Reads can be served by the replica. Writes stop until a promotion decision is taken; the control plane refuses work rather than accepting it into a database that may be about to be fenced.
A failover costs: Everything inside the replication lag at the moment of promotion. Read the lag before promoting — that number is the data-loss decision.