Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
5 contributors

Users who have contributed to this file

@pb82 @josemigallas @aliok @stianst @saibot94
168 lines (118 sloc) 8.16 KB

License

Keycloak Metrics SPI

A Service Provider that adds a metrics endpoint to Keycloak. The endpoint returns metrics data ready to be scraped by Prometheus.

Two distinct providers are defined:

  • MetricsEventListener to record the internal Keycloak events
  • MetricsEndpoint to expose the data through a custom endpoint

The endpoint lives under <url>/auth/realms/<realm>/metrics. It will return data for all realms, no matter which realm you use in the URL (you can just default to /auth/realms/master/metrics).

License

See LICENSE file

Running the tests

$ ./gradlew test

Build

The project is packaged as a jar file and bundles the prometheus client libraries.

$ ./gradlew jar

builds the jar and writes it to build/libs.

Configurable versions for some packages

You can build the project using a different version of Keycloak or Prometheus, running the command:

$ ./gradlew -PkeycloakVersion="4.7.0.Final" -PprometheusVersion="0.3.0" jar

or by changing the gradle.properties file in the root of the project.

Usage

Just drop the jar into the providers subdirectory of your Keycloak installation.

To enable the event listener via the GUI interface, go to Manage -> Events -> Config. The Event Listeners configuration should have an entry named metrics-listener.

To enable the event listener via the Keycloak CLI, such as when building a Docker container, use these commands. (These commands assume /opt/jboss is the Keycloak home directory, which is used on the jboss/keycloak reference container on Docker Hub.)

/opt/jboss/keycloak/bin/kcadm.sh config credentials --server http://localhost:8080/auth --realm master --user $KEYCLOAK_USER --password $KEYCLOAK_PASSWORD
/opt/jboss/keycloak/bin/kcadm.sh update events/config -s "eventsEnabled=true" -s "adminEventsEnabled=true" -s "eventsListeners+=metrics-listener"
/usr/bin/rm -f /opt/jboss/.keycloak/kcadm.config

Metrics

For each metric, the endpoint returns 2 or more lines of information:

  • # HELP: A small description provided by the SPI.
  • # TYPE: The type of metric, namely counter and gauge. More info about types at prometheus.io/docs.
  • Provided there were any values, the last one recorded. If no value has been recorded yet, no more lines will be given.
  • In case the same metric have different labels, there is a different line for each one. By default all metrics are labeled by realm. More info about labels at prometheus.io/docs.

Example:

# HELP jvm_memory_bytes_committed Committed (bytes) of a given JVM memory area.
# TYPE jvm_memory_bytes_committed gauge
jvm_memory_bytes_committed{area="heap",} 2.00802304E8
jvm_memory_bytes_committed{area="nonheap",} 2.0217856E8

JVM performance

A variety of JVM metrics are provided

Generic events

Every single internal Keycloak event is being shared through the endpoint, with the descriptions Generic Keycloak User event or Generic Keycloak Admin event. Most of these events are not likely useful for the majority users but are provided for good measure. A complete list of the events can be found at Keycloak documentation.

Featured events

There are however a few events that are particularly more useful from a mobile app perspective. These events have been overriden by the SPI and are described more thoroughly below.

keycloak_logins

This counter counts every login performed by a non-admin user. It also distinguishes logins by the utilised identity provider by means of the label provider and by client with the label client_id..

# HELP keycloak_logins Total successful logins
# TYPE keycloak_logins gauge
keycloak_logins{realm="test",provider="keycloak",client_id="account"} 3.0
keycloak_logins{realm="test",provider="github",client_id="application1"} 2.0
keycloak_failed_login_attempts

This counter counts every login performed by a non-admin user that fails, being the error described by the label error. It also distinguishes logins by the identity provider used by means of the label provider and by client with the label client_id.

# HELP keycloak_failed_login_attempts Total failed login attempts
# TYPE keycloak_failed_login_attempts gauge
keycloak_failed_login_attempts{realm="test",provider="keycloak",error="invalid_user_credentials",client_id="application1"} 6.0
keycloak_failed_login_attempts{realm="test",provider="keycloak",error="user_not_found",client_id="application1"} 2.0
keycloak_registrations

This counter counts every new user registration. It also distinguishes registrations by the identity provider used by means of the label provider and by client with the label client_id..

# HELP keycloak_registrations Total registered users
# TYPE keycloak_registrations gauge
keycloak_registrations{realm="test",provider="keycloak",client_id="application1"} 1.0
keycloak_registrations{realm="test",provider="github",client_id="application1"} 1.0
keycloak_registrations_errors

This counter counts every new user registration that fails, being the error described by the label error. It also distinguishes registrations by the identity provider used by means of the label provider and by client with the label client_id..

# HELP keycloak_registrations_errors Total errors on registrations
# TYPE keycloak_registrations_errors counter
keycloak_registrations_errors{realm="test",provider="keycloak",error="invalid_registration",client_id="application1",} 2.0
keycloak_registrations_errors{realm="test",provider="keycloak",error="email_in_use",client_id="application1",} 3.0
keycloak_request_duration

This histogram records the response times per route and http method and puts them in one of five buckets:

  • Requests that take 2ms or less
  • Requests that take 10ms or less
  • Requests that take 100ms or less
  • Requests that take 1s or less
  • Any request that takes longer than 1s

The response from this type of metrics has the following format:

# HELP keycloak_request_duration Request duration
# TYPE keycloak_request_duration histogram
keycloak_request_duration_bucket{method="PUT",route="/admin/realms/openshift/clients/3scale",le="2.0",} 0.0
keycloak_request_duration_bucket{method="PUT",route="/admin/realms/openshift/clients/3scale",le="10.0",} 1.0
keycloak_request_duration_bucket{method="PUT",route="/admin/realms/openshift/clients/3scale",le="100.0",} 2.0
keycloak_request_duration_bucket{method="PUT",route="/admin/realms/openshift/clients/3scale",le="1000.0",} 2.0
keycloak_request_duration_bucket{method="PUT",route="/admin/realms/openshift/clients/3scale",le="+Inf",} 2.0
keycloak_request_duration_count{method="PUT",route="/admin/realms/openshift/clients/3scale",} 2.0
keycloak_request_duration_sum{method="PUT",route="/admin/realms/openshift/clients/3scale",} 83.0

This tells you that there have been zero requests that took less than 2ms. There was one request that took less than 10ms. All the other requests took less than 100ms.

Aside from the buckets there are also the sum and count metrics for every route and method. In the above example they tell you that there have been two requests total for this route & http method. The sum of all response times for this combination is 83ms.

To get the average request duration over the last five minutes for the whole server you can use the following Prometheus query:

rate(keycloak_request_duration_sum[5m]) / rate(keycloak_request_duration_count[5m])
keycloak_response_errors

This counter counts the number of response errors (responses where the http status code is in the 400 or 500 range).

# HELP keycloak_response_errors Total number of error responses
# TYPE keycloak_response_errors counter
keycloak_response_errors{code="500",method="GET",route="/",} 1

Grafana Dashboard

You can use this dashboard or create yours https://grafana.com/dashboards/10441

You can’t perform that action at this time.