Metrics

Armada exposes metrics for Prometheus, available via the /metrics endpoint in the REST API (default port 8079). Go runtime statistics, gRPC statistics, and Dragonboat statistics, which is the underlying framework used for Raft clusters, are exposed. Some exposed metrics:

• dragonboat_raftnode_has_leader{shardid="1000",replicaid="1"} -- Raft leader for the table with shard ID 1000 is instance 1.
• grpc_server_handled_total{grpc_code="OK",grpc_method="Range",grpc_service="regatta.v1.KV",grpc_type="unary"} -- total number of served Range requests.
• regatta_table_storage_cache_hits{clusterID="10001",table="armada-test",type="block"} -- Armada table storage block cache hits
• regatta_table_storage_cache_misses{clusterID="10001",table="armada-test",type="block"} -- Armada table storage block cache misses
• regatta_table_storage_read_amp{clusterID="10001",table="armada-test"} -- Armada table storage read amplification

The gRPC service names and many metric prefixes still use the legacy regatta naming.

Alerts

Prometheus alerting rules can be found in the Helm Chart.

Debugging

Armada also exposes the /debug endpoint in the REST API for runtime profiling via pprof.