Backup & monitoring

Backup and metrics both run on the engine thread, so a snapshot is crash-consistent and metrics are exact.

• Hot consistent snapshot — engine.snapshot("./backup-DATE")?. Recover with StateMachine::open("./backup-DATE"); the recovered digest matches the live one byte-for-byte.
• Live metrics — engine.stats() returns ServerStats { applied_ops, digest, uptime_secs }. In a cluster, Node::probe() returns (digest, op_number, commit) so you can detect replica divergence by comparing across nodes.
• Prometheus — with --features http-gateway, scrape /v1/metrics for kesseldb_ops_total / kesseldb_inflight / kesseldb_view_number / kesseldb_last_op_number / kesseldb_is_primary (cardinality bounded by design).

Full reference: Usage guide (full) §11.