Module stats 

Pool, server, and peer metrics with histograms and a JSON snapshot.

The stats subsystem is split into small modules:

• Histogram - Cassandra-style estimated histogram.
• PoolField / ServerField - typed metric handles.
• Snapshot - aggregate value rendered to JSON.
• StatsServer - REST endpoint serving the latest snapshot.

Stats glues the pieces together: a writer accumulates counters, gauges, and histogram observations; a periodic aggregator publishes a fresh Snapshot that the REST endpoint serves.

Structs

Aggregator

Async aggregator handle: snapshots at a fixed interval into a shared cell that the REST server reads from.

FailureMetrics

Live, mutable accumulator for failure-cause counters.

FailureSnapshot

Immutable snapshot of every failure-cause metric.

Histogram

A fixed-bucket histogram for tracking latencies and payload sizes.

HistogramSummary

Pre-computed quantile summary derived from a Histogram.

MetricSpec

Static descriptor for a metric: how it is interpreted, what its canonical lower-case name is, and a one-line human description.

NoTargetsEntry

A single labeled dispatch_no_targets_total row.

PeerEntry

A single labeled per-peer dispatch error row.

PeerStateEntry

A single labeled peer_state_current gauge row.

PeerStats

Per-peer collected metrics. Mirrors ServerStats for cluster peers.

PhiEntry

A single labeled gossip_phi_score gauge row.

PoolStats

Per-pool collected metrics.

ServerStats

Per-datastore-server collected metrics.

ServiceInfo

Engine-wide identifying strings included in every snapshot.

Snapshot

Aggregate snapshot of the stats subsystem at a point in time.

Stats

Live, mutable counters and histograms for a single engine instance.

StatsServer

A bound TCP listener serving the stats endpoint.

TimeoutEntry

A single labeled dispatch_response_timeout_total row.

TransitionEntry

A single labeled peer_state_transitions_total row.

Enums

Latency

Channels used to mutate histogram observations.

PoolField

Typed handle for a pool metric.

QueueGauge

Channels used for queue-length observations (observed at sample time, not events).

QueueWait

Channels used for queue-wait-time observations.

ServerField

Typed handle for a server metric.

StatsMetricType

Kind of metric tracked by the stats subsystem.

Constants

BUCKET_COUNT

Number of buckets in the estimated histogram.

MAX_HEADERS

Maximum number of headers parsed in a single request.

MAX_REQUEST_BYTES

Maximum number of bytes the server will read for an HTTP request line plus headers. Requests larger than this are rejected.

POOL_CODEC

Const slice of every pool metric descriptor in declaration order.

SERVER_CODEC

Const slice of every server metric descriptor in declaration order.

Functions

describe_stats

Returns the human-readable description block printed by the -D command-line flag.

render_prometheus

Render a Snapshot in the Prometheus 0.0.4 text exposition format.

Type Aliases

ClusterInfoProvider

Type alias for a closure that produces a fresh ClusterInfoSnapshot every time the /cluster-info.txt route is hit.