Slap.Streams.Metrics (slap_streams v0.4.0)

Prometheus metrics from Slap.Streams.Telemetry's events and slap_cluster's, kept in ETS and rendered in the text exposition format by render/0 (served by Slap.Streams.HTTP.Metrics).

One metrics process runs per VM. Its labels do not distinguish separate Streams clusters on that VM. It ignores cluster events from other services, such as Slap.KV.Cluster.

Metric	Type	Labels
`slap_streams_append_duration_seconds`	histogram	accept to durable reply
`slap_streams_appends_total`	counter	`result`
`slap_streams_append_bytes_total`	counter
`slap_streams_appends_rejected_total`	counter	`reason` (backpressure)
`slap_streams_write_failures_total`	counter
`slap_streams_stream_servers`	gauge	`shard`
`slap_streams_inflight_bytes`, `slap_streams_inflight_requests`	gauge	`shard`
`slap_streams_waiters`	gauge	`shard` (long-poll and SSE)
`slap_streams_delete_backlog`, `slap_streams_expiry_due`	gauge	`shard` (job backlogs)
`slap_streams_l0_ssts`, `slap_streams_sorted_runs`	gauge	`shard` (compaction backlog)
`slap_streams_block_cache_hits_total`, `slap_streams_block_cache_misses_total`	counter	`shard`
`slap_streams_durability_lag`	gauge	`shard` (seqs written, not durable)
`slap_streams_shard_events_total`	counter	`event` (`start`, `stop`, `fenced`)
`slap_streams_store_probe_ok`	gauge	1 if the most recent store probe passed
`slap_streams_http_requests_total`	counter	`method`, `status`
`slap_streams_http_request_duration_seconds`	histogram	`method`

This is a small exporter of its own rather than telemetry_metrics and its Prometheus reporter: the metrics are fixed, and it keeps the dependencies to telemetry.

Summary

Functions

child_spec(init_arg)

Returns a specification to start this module under a supervisor.

render()

The metrics in the Prometheus text format.

start_link(opts \\ [])

Functions

child_spec(init_arg)

Returns a specification to start this module under a supervisor.

See Supervisor.

render()

@spec render() :: iodata()

The metrics in the Prometheus text format.