Skip to content

Latest commit

 

History

History
147 lines (102 loc) · 7.42 KB

README.md

File metadata and controls

147 lines (102 loc) · 7.42 KB
Raw

Application Monitoring Operator

An OpenShift Operator using the Operator SDK that installs an Application Monitoring Stack consisting of Grafana, Prometheus & Alertmanager.

Warning:

This operator isn't being actively maintained, issues may not be resolved unless they are security related.

Supported Custom Resources

The following resources are supported:

ApplicationMonitoring

Triggers the installation of the monitoring stack when created. This is achieved by deploying two other operators that install Prometheus and Grafana respectively.

The Application Monitoring CR accepts the following properties in the spec:

BlackboxTarget

The Blackbox Target CR accepts the following properties in the spec:

  • blackboxTargets: A list of targets for the blackbox exporter to probe.

The blackboxTargets should be provided as an array in the form of:

  blackboxTargets:
    - service: example
      url: https://example.com
      module: http_extern_2xx

where service will be added as a label to the metric, url is the URL of the route to probe and module can be one of:

  • http_2xx: Probe http or https targets via GET using the cluster certificates
  • http_post_2xx: Probe http or https targets via POST using the cluster certificates
  • http_extern_2xx: Probe http or https targets via GET relying on a valid external certificate

PrometheusRule

Represents a set of alert rules for Prometheus/Alertmanager. See the prometheus operator docs for more details about this resource. An example PrometheusRule can be seen in the example app template

ServiceMonitor

Represents a Service to pull metrics from. See the prometheus operator docs for more details about this resource. An example ServiceMonitor can be seen in the example app template

PodMonitor

Represents pods to pull metrics from. See the prometheus operator docs for more details about this resource.

GrafanaDashboard

Represents a Grafana dashboard. You typically create this in the namespace of the service the dashboard is associated with. The Grafana operator reconciles this resource into a dashboard. An example GrafanaDashboard can be seen in the example app template

Installation

You will need cluster admin permissions to create CRDs, ClusterRoles & ClusterRoleBindings. ClusterRoles are needed to allow the operators to watch multiple namespaces.

make cluster/install

You can access Grafana, Prometheus & AlertManager web consoles using the Routes in the project.

Verify installation

Run the following commands

# Check the project exists
$ oc project
Using project "application-monitoring" on server "https://your-cluster-ip:8443"

# Check the pods exist e.g.
$ oc get pods
NAME                                              READY     STATUS    RESTARTS   AGE
alertmanager-application-monitoring-0             2/2       Running   0          1h
application-monitoring-operator-77cdbcbff-fbrnr   1/1       Running   0          1h
grafana-deployment-6dc8df6bb4-rxdjs               1/1       Running   0          49m
grafana-operator-7c4869cfdc-6sdv9                 1/1       Running   0          1h
prometheus-application-monitoring-0               4/4       Running   1          36m
prometheus-operator-7547bb757b-46lwh              1/1       Running   0          1h

Example Monitored Project

These steps create a new project with a simple server exposing a metrics endpoint. The template also includes a simple PrometheusRule, ServiceMonitor & GrafanaDashboard custom resource that the application-monitoring stack detects and reconciles.

oc new-project example-prometheus-nodejs
oc label namespace example-prometheus-nodejs monitoring-key=middleware
oc process -f https://raw.githubusercontent.com/david-martin/example-prometheus-nodejs/master/template.yaml | oc create -f -

You should see the following things once reconciled:

  • New Grafana Dashboard showing memory usage of the app (among other things)
  • A new Target example-prometheus-nodejs/example-prometheus-nodejs in Prometheus
  • A new Alert Rule APIHighMedianResponseTime

The example application provides three endpoints that will produce more metrics:

  • / will return Hello World after a random response time
  • /checkout Will create random checkout metrics
  • /bad Can be used to create error rate metrics

Running locally (for development)

You can run the Operator locally against a remote namespace. The name of the namespace should be application-monitoring. To run the operator execute:

$ make setup/gomod
$ make cluster/install/local
$ make code/run

Generating CSV

  1. Update the AMO_VERSION and PREV_AMO_VERSION values in Makefile
  2. Run make gen/csv to generate a new manifest.
  3. Ensure clusterPermissions block in generated csv is up to date with cluster roles in cluster-roles directory.

Ensure:

  • The new csv file points to the latest version of the operator image. Note the images are referenced twice in the csv.
  • deploy/operator.yaml has the correct image version tag. All image tags should be prefixed with a v
  • application-monitoring-operator.package.yaml references the correct version.

Release

  1. Update the version number (e.g. 1.0.1) in the following files: Makefile, operator.yaml, version.go)

  2. Create a pull request with the updated version. Once it has been reviewed and approved, merge it to master

  3. Create a new release via the GitHub release page, naming it the same as the version number - e.g. 1.0.1. Ensure to state what's new in the release

  4. Login to quay.io: docker login quay.io

  5. Build the Docker image for the new version, and push it to quay.io:

    $ make image/build
$ make image/push