Status and monitoring

What each component exposes today. Nothing here needs the hosted service; the LokaDB Cloud panel shows the same state with statistics and alerts.

The proxy

Set health_listen in the proxy config to get two internal endpoints (do not expose them outside your network):

• GET /healthz: 200, or 503 while the proxy drains before a stop;
• GET /sessions: the open sessions.

The proxy's logs never contain values from data rows or passwords, and SQL in them has its literals replaced by $n.

The relay

GET /healthz on the relay answers 200, and 503 as soon as it is asked to stop; use it as the readiness probe behind a load balancer or in Kubernetes (the tunnel in production). lokadb agents -cp DSN lists the enrolled agents from the registry.

The agent

Start the agent with -status-listen ADDR to serve GET /status locally: its state, the relays it is connected to, its certificate, the time to its -db. It reports counters only, no data. With -status-url URL and -status-token-file FILE it also posts the same status to a LokaDB service every 60 seconds; set the environment variable LOKADB_AGENT_STATUS=off to stop that.

A move in progress

move-org logs its progress (rows copied, rounds applied, the pause of the switch) and check reports leftovers of a move that was killed (lokadb move-cleanup removes them). See enable and move.