Skip to content
v0.5.1GitHub

Using

`GET /metrics` — system + all domain metrics. Admin only.

applies to v0.5.1
GET
/store-api/metrics
curl --request GET \
--url https://example.com/store-api/metrics \
--header 'Authorization: Bearer <token>'

engines is an additive block, one entry per storage engine (kv/json/rel), each with the same nine fields: read_ops, write_ops, read_latency_us_p50/p95/p99, write_latency_us_p50/p95/p99, window_secs. Clients compute read_ops / window_secs for ops/s — the rate is a window_secs-wide average, not an instantaneous value, since read_ops/write_ops only count fully ticked seconds (the running second is excluded). A disabled engine still gets its block, all zero — 0 for a latency percentile means no op landed in the window, not an unmeasurably fast one.

shm covers the shared-memory snapshot publisher, all zero while SHM is off: snapshot_publishes_total, snapshot_publish_skips_total (a client was still reading the buffer due for overwriting), snapshot_consecutive_skips (the current run of skips; above 0 the snapshot is not being updated) and clients_kicked_total (clients disconnected after shm.kick_blocking_client_after_secs).

resources is the full disk and memory report, of which GET /health gives the states only: per data volume path, size_bytes, free_bytes and state, and for the memory rss_anon_bytes, budget_bytes and state. io appears only while writes of a data file fail: the file with the most failures in a row, path and failures.

Metrics snapshot

Unauthorized

Media typetext/plain
string

Forbidden — Admin only

Media typetext/plain
string

A data shard is not running

Media typetext/plain
string