Skip to content

METRICS.md

Latest commit

 

History

History
43 lines (32 loc) · 3.22 KB

METRICS.md

File metadata and controls

43 lines (32 loc) · 3.22 KB

Metrics Reference

This document describes each metric emitted by the library via the metrics.Recorder interface. The default implementation is a no‑op; you can supply a concrete recorder (e.g. the Prometheus example under examples/prometheus) using impl.WithMetricsRecorder.

Method Purpose Labels / Notes
RecordStartupRecovery(success bool, unackedCount int, duration time.Duration) Invoked during initialization when the client recovers unacked messages from the LBS. success = true/false. Duration measured from start to completion. unackedCount is informational only.
RecordClaimAttempt(streamName string, success bool, duration time.Duration) A consumer attempted to claim an expired stream after receiving a StreamExpired notification. Labels: stream, success.
RecordLockAcquisitionAttempt(streamName string, success bool, duration time.Duration) Internal lock acquisition (including recovery on startup and when taking possession after a claim). Labels: stream, success.
RecordLockExtensionAttempt(streamName string, success bool) A heartbeat attempt to extend an existing lock. Labels: stream, success.
RecordLockReleaseAttempt(streamName string, success bool) Unlocking a stream (typically in DoneStream or when a lock expires). Labels: stream, success.
RecordStreamProcessingStart(streamName string, startTime time.Time) Called when a stream is assigned and processing begins. Used in combination with RecordStreamProcessingEnd to compute duration.
RecordStreamProcessingEnd(streamName string, endTime time.Time) Called when processing for a stream finishes. Duration is endTime - startTime.
RecordKspNotification(streamName string) A keyspace notification (expiration) was received for a stream. Label: stream.
RecordKspNotificationDropped() A keyspace notification was dropped due to a full internal channel. No labels.

Using with Prometheus

See examples/prometheus/recorder.go for a reference implementation. To use it:

import (
    prom "github.com/handcoding-labs/redis-stream-client-go/examples/prometheus"
    "github.com/prometheus/client_golang/prometheus/promhttp"
)

rec := prom.NewPrometheusRecorder(prometheus.DefaultRegisterer)
client, _ := impl.NewRedisStreamClient(redisClient, "my-service", impl.WithMetricsRecorder(rec))
// …start HTTP server with promhttp.Handler() …

The example registers counters and histograms – feel free to copy and adapt to your own monitoring system.

Significance

  • Startup recovery metrics help you understand how many unprocessed streams your service recovers when it restarts. High durations may indicate Redis latency or large pending queues.
  • Claim / lock metrics provide visibility into contention between consumers and lock collisions.
  • Stream processing durations quantify end-to-end time per stream (useful for SLA monitoring).
  • Keyspace notification counts show the volume of consumer expirations, which may correlate with crash rates or slow consumers.
  • A non‑zero KspNotificationDropped indicates backpressure in the library; consider increasing WithKspChanSize or processing notifications faster.

"```