Monitoring & Alerting

Goose emits metrics around Job execution timings, success rates, latency & queue sizes. A Metric plugin receives these metrics & forwards it to its respective backend. Goose provides StatsD as a specimen metric backend.

For Prometheus, or other metrics backends, a plugin can be injected by following this guide.

List of all Metrics

Metric	Type	Description
enqueued_jobs.<my-queue>.count	gauge	Count of jobs in my-queue
total_enqueued_jobs.count	gauge	Total count of all jobs
scheduled_jobs.count	gauge	Count of scheduled jobs
batches.count	gauge	Count of batches
periodic_jobs.count	gauge	Count of cron jobs
dead_jobs.count	gauge	Count of dead jobs
jobs.processed	count	Count of processed jobs
jobs.succeeded	count	Count of successful jobs
jobs.failed	count	Count of failed jobs
jobs.recovered	count	Count of orphan jobs which were recovered
batch.success	count	Count of batches where all jobs succeeded
batch.dead	count	Count of batches where all jobs died
batch.partial-success	count	Count of batches with mix of successful and dead jobs
execution.latency	timing	Latency(in ms) between enqueue -> start of execution
scheduled.latency	timing	Latency(in ms) between theoretical schedule time -> start execution
cron_scheduled.latency	timing	Latency(in ms) between theoretical schedule time -> start execution
retry.latency	timing	Latency(in ms) between theoretical retry time -> start of execution
job.execution_time	timing	Time taken to execute a job(in ms)
batch.completion_time	timing	Time taken to complete a batch(in ms)

Configuring StatsD

StatsD plugin an be configured in following ways:

Key	Description
:enabled?	Boolean flag for enabling/disabling metrics
:host	Host of StatsD Aggregator
:port	Port of StatsD Aggregator
:prefix	Prefix for all metrics. Can be a generic term like "goose." or specific to microservice's name
:sample-rate	Sample rate of metric collection
:tags	Map of key-value pairs to be attached to every metric

Usage

Note: Goose uses clj-statsd, which uses agents internally. Post stopping a worker, (shutdown-agents) must be called in order to exit the program.

(ns statsd-metrics
  (:require
    [goose.metrics.statsd :as statsd]
    [goose.worker :as w]))

(let [statsd-opts {:enabled?    true
                   :host        "localhost"
                   :port        8125
                   :prefix      "maverick."
                   :sample-rate 0.9
                   :tags        {:top :gun}}
      statsd (statsd/new statsd-opts)
      worker-opts (assoc worker-opts :metrics-plugin statsd)
      worker (w/start worker-opts)]
  ;; When shutting down worker...
  (w/stop worker)
  ;; clj-statsd uses agents internally. Call (shutdown-agents) to exit the program.
  (shutdown-agents))

---

Previous: Error Handling & Retries        Next: Guide to Custom Metrics Backend