iapetos

iapetos is a Clojure wrapper around the Prometheus Java Client, providing idiomatic and simple access to commonly used functionality while retaining low-level flexibility for tackling more complex tasks.

N.B. Since version 0.1.9, iapetos is released as clj-commons/iapetos on Clojars. Previously it was available as xsc/iapetos.

Table of Contents

• Basic Usage
  • Registering Metrics
  • Metric Export
  • Metric Push
  • Labels
  • Subsystems
• Features
  • Code Block Instrumentation
  • JVM Metrics
  • Function Instrumentation
  • Ring
    • Exception Handling
  • Standalone HTTP Server
• History
• License

Basic Usage

Registering Metrics

Documentation

All metrics have to be registered with a collector registry before being used:

(require '[iapetos.core :as prometheus])

(defonce registry
  (-> (prometheus/collector-registry)
      (prometheus/register
        (prometheus/histogram :app/duration-seconds)
        (prometheus/gauge     :app/active-users-total)
        (prometheus/counter   :app/runs-total))
      (prometheus/register-lazy
        (prometheus/gauge     :app/last-success-unixtime))))

Now, they are ready to be set and changed:

(-> registry
    (prometheus/inc     :app/runs-total)
    (prometheus/observe :app/duration-seconds 0.7)
    (prometheus/set     :app/active-users-total 22))

The registry itself implements clojure.lang.IFn to allow access to all registered metrics (plus setting of metric labels), e.g.:

(registry :app/duration-seconds)
;; => #object[io.prometheus.client.Histogram$Child ...]

All metric operations can be called directly on such a collector, i.e.:

(prometheus/inc     (registry :app/runs-total))
(prometheus/observe (registry :app/duration-seconds) 0.7)
(prometheus/set     (registry :app/active-users-total) 22)

Metric Export

Documentation

Metrics can be transformed into a textual representation using iapetos.export/text-format:

(require '[iapetos.export :as export])

(print (export/text-format registry))
;; # HELP app_active_users_total a gauge metric.
;; # TYPE app_active_users_total gauge
;; app_active_users_total 22.0
;; # HELP app_runs_total a counter metric.
;; # TYPE app_runs_total counter
;; app_runs_total 1.0
;; ...

This could now be exposed e.g. using an HTTP endpoint (see also iapetos' Ring integration or the standalone server ).

Metric Push

Documentation

Another way of communicating metrics to Prometheus is using push mechanics, intended to be used for e.g. batch jobs that might not live long enough to be scraped in time. Iapetos offers a special kind of registry for this:

(require '[iapetos.export :as export])

(defonce registry
  (-> (export/pushable-collector-registry
        {:push-gateway "push-gateway-host:12345"
         :job          "my-batch-job"})
      (prometheus/register ...)))
...
(export/push! registry)

Note that you can reduce the amount of boilerplate in most cases down to something like:

(export/with-push-gateway [registry {:push-gateway "...", :job "..."}]
  (-> registry
      (prometheus/register
        (prometheus/counter :app/rows-inserted-total)
        ...)
      (run-job! ...)))

Labels

Prometheus allows for labels to be associated with metrics which can be declared for each collector before it is registered:

(def job-latency-histogram
  (prometheus/histogram
    :app/job-latency-seconds
    {:description "job execution latency by job type"
     :labels [:job-type]
     :buckets [1.0 5.0 7.5 10.0 12.5 15.0]}))

(defonce registry
  (-> (prometheus/collector-registry)
      (prometheus/register job-latency-histogram)))

Now, you can lookup a collector bound to a set of labels by calling the registry with a label/value-map:

(prometheus/observe (registry :app/job-latency-seconds {:job-type "pull"}) 14.2)
(prometheus/observe (registry :app/job-latency-seconds {:job-type "push"}) 8.7)

(print (export/text-format registry))
;; # HELP app_job_latency_seconds job execution latency by job type
;; # TYPE app_job_latency_seconds histogram
;; app_job_latency_seconds_bucket{job_type="pull",le="1.0",} 0.0
;; app_job_latency_seconds_bucket{job_type="pull",le="5.0",} 0.0
;; ...
;; app_job_latency_seconds_bucket{job_type="push",le="1.0",} 0.0
;; app_job_latency_seconds_bucket{job_type="push",le="5.0",} 0.0
;; ...

Subsystems

In addition to namespaces, you can create collector declarations belonging to a subsystem, i.e.:

(prometheus/counter
  :app/job-runs-total
  {:description "the total number of finished job executions."
   :subsystem "worker"})

But this reduces its reusability - you might want to register the above counter twice in different subsystems without having to create it anew - which is why iapetos lets you specify the subsystem on the registry level:

(defonce registry
  (prometheus/collector-registry))

(defonce worker-registry
  (-> registry
      (prometheus/subsystem "worker")
      (prometheus/register ...)))

(defonce httpd-registry
  (-> registry
      (prometheus/subsystem "httpd")
      (prometheus/register ...)))

Now, collectors added to worker-registry and httpd-registry will have the appropriate subsystem. And when registry is exported it will contain all metrics that were added to the subsystems.

(Note, however, that the subsystem registries will not have access to the original registry's collectors, i.e. you have to reregister things like function instrumentation or Ring collectors.)

Features

Code Block Instrumentation

Documentation

iapetos provides a number of macros that you can use to instrument parts of your code, e.g. with-failure-timestamp to record the last time a task has thrown an error or with-duration to track execution time:

(prometheus/with-failure-timestamp (registry :app/last-worker-failure-unixtime)
  (prometheus/with-duration (registry :app/worker-latency-seconds)
    (run-worker! task)))

See the auto-generated documentation for all available macros or the function instrumentation section below on how to easily wrap them around existing functions.

JVM Metrics

Documentation

Some characteristics of your current JVM are always useful (e.g. memory usage, thread count, ...) and can be added to your registry using the iapetos.collector.jvm namespace:

(require '[iapetos.collector.jvm :as jvm])

(defonce registry
  (-> (prometheus/collector-registry)
      (jvm/initialize)))

Alternatively, you can selectively register the JVM collectors:

(defonce registry
  (-> (prometheus/collector-registry)
      (prometheus/register
        (jvm/standard)
        (jvm/gc)
        (jvm/memory-pools)
        (jvm/threads))))

Note: You need to include the artifact io.prometheus/simpleclient_hotspot explicitly in your project's dependencies.

Function Instrumentation

Documentation

To collect metrics about specific functions, you can use the functionality provided in iapetos.collector.fn:

(require '[iapetos.collector.fn :as fn])

(defn- run-the-job!
  [job]
  ...)

(defonce registry
  (-> (prometheus/collector-registry)
      ...
      (fn/initialize)))

(fn/instrument! registry #'run-the-job!)

Now, every call to run-the-job! will update a series of duration, success and failure metrics. Note, however, that re-evaluation of the run-the-job! declaration will remove the instrumentation again.

Ring

Documentation

iapetos.collector.ring offers middlewares to

• expose a iapetos collector registry via a fixed HTTP endpoint, and
• collect metrics for Ring handlers.

First, you need to initialize the available collectors in the registry:

(require '[iapetos.collector.ring :as ring])

(defonce registry
  (-> (prometheus/collector-registry)
      (ring/initialize)))

Afterwards, you can add the middlewares to your Ring stack:

(def app
  (-> (constantly {:status 200})
      (ring/wrap-metrics registry {:path "/metrics"})))

The following metrics will now be collected and exposed via the GET /metrics endpoint:

• http_requests_total
• http_request_latency_seconds
• http_exceptions_total

These are, purposefully, compatible with the metrics produced by prometheus-clj, as to allow a smooth migration.

Ring supports sync and async handlers, metrics are collected for both by iapetos.collector.ring ns.

Exception Handling

By default, if your ring handler throws an exception, only the http_exceptions_total counter would be incremented. This means that if you respond with a 500 error code on exceptions:

1. These responses won't be counted on http_requests_total
2. Their latencies won't be observed on http_request_latency_seconds

To overcome this, you can use the optional :exception-status to define a status code to be reported on both metrics, for example:

(def app
  (-> (fn [_] (throw (Exception.)))
      (ring/wrap-metrics registry {:path "/metrics" :exception-status 500})))

will increment all 3 metrics, assuming a 500 response code for exceptions:

• http_requests_total
• http_request_latency_seconds
• http_exceptions_total

Standalone HTTP Server

Documentation

A zero-dependency standalone HTTP server is included in iapetos.standalone and can be run using:

(require '[iapetos.standalone :as standalone])

(defonce httpd
  (standalone/metrics-server registry {:port 8080}))

This is particularly useful for applications that do not expose an HTTP port themselves but shall still be scraped by Prometheus. By default, metrics will be exposed at /metrics.

History

iapetos was originally created by Yannick Scherer (@xsc). In July 2019 it was moved to CLJ Commons for continued maintenance.

It could previously be found at xsc/iapetos. clj-commons/iapetos is the canonical repository now.

License

MIT License

Copyright (c) 2016 Yannick Scherer

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.