Design

This document describes the design and interaction between the custom resource definitions that the Prometheus Operator introduces.

The custom resources that the Prometheus Operator introduces are:

  • Prometheus
  • ServiceMonitor
  • Alertmanager

Prometheus

The Prometheus custom resource definition (CRD) declaratively defines a desired Prometheus setup to run in a Kubernetes cluster. It provides options to configure replication, persistent storage, and Alertmanagers to which the deployed Prometheus instances send alerts to.

For each Prometheus resource, the Operator deploys a properly configured StatefulSet in the same namespace. The Prometheus Pods are configured to mount a Secret called <prometheus-name> containing the configuration for Prometheus.

The CRD specifies which ServiceMonitors should be covered by the deployed Prometheus instances based on label selection. The Operator then generates a configuration based on the included ServiceMonitors and updates it in the Secret containing the configuration. It continuously does so for all changes that are made to ServiceMonitors or the Prometheus resource itself.

If no selection of ServiceMonitors is provided, the Operator leaves management of the Secret to the user, which allows to provide custom configurations while still benefiting from the Operator's capabilities of managing Prometheus setups.

ServiceMonitor

The ServiceMonitor custom resource definition (CRD) allows to declaratively define how a dynamic set of services should be monitored. Which services are selected to be monitored with the desired configuration is defined using label selections. This allows an organization to introduce conventions around how metrics are exposed, and then following these conventions new services are automatically discovered, without the need to reconfigure the system.

For Prometheus to monitor any application within Kubernetes an Endpoints object needs to exist. Endpoints objects are essentially lists of IP addresses. Typically an Endpoints object is populated by a Service object. A Service object discovers Pods by a label selector and adds those to the Endpoints object.

A Service may expose one or more service ports, which are backed by a list of multiple endpoints that point to a Pod in the common case. This is reflected in the respective Endpoints object as well.

The ServiceMonitor object introduced by the Prometheus Operator in turn discovers those Endpoints objects and configures Prometheus to monitor those Pods.

The endpoints section of the ServiceMonitorSpec, is used to configure which ports of these Endpoints are going to be scraped for metrics, and with which parameters. For advanced use cases one may want to monitor ports of backing Pods, which are not directly part of the service endpoints. Therefore when specifying an endpoint in the endpoints section, they are strictly used.

Note: endpoints (lowercase) is the field in the ServiceMonitor CRD, while Endpoints (capitalized) is the Kubernetes object kind.

While ServiceMonitors must live in the same namespace as the Prometheus resource, discovered targets may come from any namespace. This is important to allow cross-namespace monitoring use cases, e.g. for meta-monitoring. Using the namespaceSelector of the ServiceMonitorSpec, one can restrict the namespaces the Endpoints objects are allowed to be discovered from. To discover targets in all namespaces the namespaceSelector has to be empty:

spec:
  namespaceSelector:
    any: true

Alertmanager

The Alertmanager custom resource definition (CRD) declaratively defines a desired Alertmanager setup to run in a Kubernetes cluster. It provides options to configure replication and persistent storage.

For each Alertmanager resource, the Operator deploys a properly configured StatefulSet in the same namespace. The Alertmanager pods are configured to include a Secret called <alertmanager-name> which holds the used configuration file in the key alertmanager.yaml.

When there are two or more configured replicas the operator runs the Alertmanager instances in high availability mode.