Prometheus metrics

This page lists every series on /metrics, from internal/metrics. The endpoint is mounted on the ops listener when features.enable_metrics is on (the default) and serves the server’s own registry, not the Prometheus global one. No label carries a peer ID.

Requests

Recorded for every request on the seven protocol pipelines.

Name	Type	Labels	Meaning
ricochet_requests_total	counter	protocol, operation, outcome	Protocol requests by outcome
ricochet_request_duration_seconds	histogram	protocol, operation, outcome	Request latency; buckets 0.001, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10 s

Label	Values
protocol	msa, msa_batch, maa, mma, sda, sfa, sca
operation	submit (MSA), batch_submit (MSA batch), the operation name for the others (such as retrieve, createMailbox, PUT, APPEND, QUERY), or unrouted when the request named no known operation
outcome	ok, client_error (4xx in the response), server_error (5xx in the response), rate_limited (a per-peer rate limit refused it), overloaded (admission control shed it, or the server is draining), error (decode failure, unroutable operation, panic)

The operations per protocol are on Protocol IDs.

Admission control

Name	Type	Labels	Meaning
ricochet_requests_in_flight	gauge	none	Requests holding an admission slot
ricochet_admission_max_in_flight	gauge	none	Server-wide concurrent request bound
ricochet_admission_admitted_total	counter	none	Requests admitted since start
ricochet_admission_shed_total	counter	none	Requests shed because the server was at capacity
ricochet_admission_waited_total	counter	none	Requests that waited for a slot
ricochet_admission_tracked_peers	gauge	none	Peers holding or awaiting a slot

Database pool

Registered when the storage backend is PostgreSQL.

Name	Type	Labels	Meaning
ricochet_db_pool_conns_total	gauge	none	Connections in the pool, idle and in use
ricochet_db_pool_conns_idle	gauge	none	Idle connections
ricochet_db_pool_conns_acquired	gauge	none	Connections checked out
ricochet_db_pool_conns_max	gauge	none	Configured pool size
ricochet_db_pool_empty_acquire_total	counter	none	Acquires that found no free connection and waited
ricochet_db_pool_canceled_acquire_total	counter	none	Acquires abandoned before a connection became free
ricochet_db_pool_acquire_seconds_total	counter	none	Cumulative time spent waiting to acquire a connection

Frame buffer pool

Name	Type	Labels	Meaning
ricochet_buffer_pool_hits_total	counter	none	Frame buffers served from the pool
ricochet_buffer_pool_misses_total	counter	none	Frame buffers that had to be allocated

Storage capacity

Taken from the capacity sample, which runs at startup and every intervals.cleanup_min. These series are absent until the first sample completes.

Name	Type	Labels	Meaning
ricochet_mailboxes	gauge	none	Mailboxes across all owners
ricochet_messages_stored	gauge	none	Messages stored
ricochet_message_bytes	gauge	none	Summed message payload length
ricochet_database_bytes	gauge	none	Database size on disk, including indexes and unreclaimed space
ricochet_storage_max_bytes	gauge	none	Configured storage budget (storage.max_storage_gb)
ricochet_mailboxes_near_capacity	gauge	none	Mailboxes at or above storage.near_capacity_ratio of their cap
ricochet_mailbox_depth	gauge	bucket	Mailboxes by message count; bucket is 0, 1-9, 10-99, 100-999, 1000-9999 or 10000+
ricochet_stats_age_seconds	gauge	none	Age of the sample these figures came from

Go runtime and process

The registry includes the standard collectors from prometheus/client_golang: the Go collector (go_* series) and the process collector (process_* series).

See Metrics & the ops view.