Chapter 2. Cluster Observability Operator overview


Important

The Cluster Observability Operator is a Technology Preview feature only. Technology Preview features are not supported with Red Hat production service level agreements (SLAs) and might not be functionally complete. Red Hat does not recommend using them in production. These features provide early access to upcoming product features, enabling customers to test functionality and provide feedback during the development process.

For more information about the support scope of Red Hat Technology Preview features, see Technology Preview Features Support Scope.

The Cluster Observability Operator (COO) is an optional component of the OpenShift Container Platform designed for creating and managing highly customizable monitoring stacks. It enables cluster administrators to automate configuration and management of monitoring needs extensively, offering a more tailored and detailed view of each namespace compared to the default OpenShift Container Platform monitoring system.

The COO deploys the following monitoring components:

  • Prometheus - A highly available Prometheus instance capable of sending metrics to an external endpoint by using remote write.
  • Thanos Querier (optional) - Enables querying of Prometheus instances from a central location.
  • Alertmanager (optional) - Provides alert configuration capabilities for different services.
  • UI plugins (optional) - Enhances the observability capabilities with plugins for monitoring, logging, distributed tracing and troubleshooting.
  • Korrel8r (optional) - Provides observability signal correlation, powered by the open source Korrel8r project.

2.1. COO compared to default monitoring stack

The COO components function independently of the default in-cluster monitoring stack, which is deployed and managed by the Cluster Monitoring Operator (CMO). Monitoring stacks deployed by the two Operators do not conflict. You can use a COO monitoring stack in addition to the default platform monitoring components deployed by the CMO.

The key differences between COO and the default in-cluster monitoring stack are shown in the following table:

FeatureCOODefault monitoring stack

Scope and integration

Offers comprehensive monitoring and analytics for enterprise-level needs, covering cluster and workload performance.

However, it lacks direct integration with OpenShift Container Platform and typically requires an external Grafana instance for dashboards.

Limited to core components within the cluster, for example, API server and etcd, and to OpenShift-specific namespaces.

There is deep integration into OpenShift Container Platform including console dashboards and alert management in the console.

Configuration and customization

Broader configuration options including data retention periods, storage methods, and collected data types.

The COO can delegate ownership of single configurable fields in custom resources to users by using Server-Side Apply (SSA), which enhances customization.

Built-in configurations with limited customization options.

Data retention and storage

Long-term data retention, supporting historical analysis and capacity planning

Shorter data retention times, focusing on short-term monitoring and real-time detection.

2.2. Key advantages of using COO

Deploying COO helps you address monitoring requirements that are hard to achieve using the default monitoring stack.

2.2.1. Extensibility

  • You can add more metrics to a COO-deployed monitoring stack, which is not possible with core platform monitoring without losing support.
  • You can receive cluster-specific metrics from core platform monitoring through federation.
  • COO supports advanced monitoring scenarios like trend forecasting and anomaly detection.

2.2.2. Multi-tenancy support

  • You can create monitoring stacks per user namespace.
  • You can deploy multiple stacks per namespace or a single stack for multiple namespaces.
  • COO enables independent configuration of alerts and receivers for different teams.

2.2.3. Scalability

  • Supports multiple monitoring stacks on a single cluster.
  • Enables monitoring of large clusters through manual sharding.
  • Addresses cases where metrics exceed the capabilities of a single Prometheus instance.

2.2.4. Flexibility

  • Decoupled from OpenShift Container Platform release cycles.
  • Faster release iterations and rapid response to changing requirements.
  • Independent management of alerting rules.

2.3. Target users for COO

COO is ideal for users who need high customizability, scalability, and long-term data retention, especially in complex, multi-tenant enterprise environments.

2.3.1. Enterprise-level users and administrators

Enterprise users require in-depth monitoring capabilities for OpenShift Container Platform clusters, including advanced performance analysis, long-term data retention, trend forecasting, and historical analysis. These features help enterprises better understand resource usage, prevent performance issues, and optimize resource allocation.

2.3.2. Operations teams in multi-tenant environments

With multi-tenancy support, COO allows different teams to configure monitoring views for their projects and applications, making it suitable for teams with flexible monitoring needs.

2.3.3. Development and operations teams

COO provides fine-grained monitoring and customizable observability views for in-depth troubleshooting, anomaly detection, and performance tuning during development and operations.

2.4. Using Server-Side Apply to customize Prometheus resources

Server-Side Apply is a feature that enables collaborative management of Kubernetes resources. The control plane tracks how different users and controllers manage fields within a Kubernetes object. It introduces the concept of field managers and tracks ownership of fields. This centralized control provides conflict detection and resolution, and reduces the risk of unintended overwrites.

Compared to Client-Side Apply, it is more declarative, and tracks field management instead of last applied state.

Server-Side Apply
Declarative configuration management by updating a resource’s state without needing to delete and recreate it.
Field management
Users can specify which fields of a resource they want to update, without affecting the other fields.
Managed fields
Kubernetes stores metadata about who manages each field of an object in the managedFields field within metadata.
Conflicts
If multiple managers try to modify the same field, a conflict occurs. The applier can choose to overwrite, relinquish control, or share management.
Merge strategy
Server-Side Apply merges fields based on the actor who manages them.

Procedure

  1. Add a MonitoringStack resource using the following configuration:

    Example MonitoringStack object

    apiVersion: monitoring.rhobs/v1alpha1
    kind: MonitoringStack
    metadata:
      labels:
        coo: example
      name: sample-monitoring-stack
      namespace: coo-demo
    spec:
      logLevel: debug
      retention: 1d
      resourceSelector:
        matchLabels:
          app: demo

  2. A Prometheus resource named sample-monitoring-stack is generated in the coo-demo namespace. Retrieve the managed fields of the generated Prometheus resource by running the following command:

    $ oc -n coo-demo get Prometheus.monitoring.rhobs -oyaml --show-managed-fields

    Example output

    managedFields:
    - apiVersion: monitoring.rhobs/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:labels:
            f:app.kubernetes.io/managed-by: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
          f:ownerReferences:
            k:{"uid":"81da0d9a-61aa-4df3-affc-71015bcbde5a"}: {}
        f:spec:
          f:additionalScrapeConfigs: {}
          f:affinity:
            f:podAntiAffinity:
              f:requiredDuringSchedulingIgnoredDuringExecution: {}
          f:alerting:
            f:alertmanagers: {}
          f:arbitraryFSAccessThroughSMs: {}
          f:logLevel: {}
          f:podMetadata:
            f:labels:
              f:app.kubernetes.io/component: {}
              f:app.kubernetes.io/part-of: {}
          f:podMonitorSelector: {}
          f:replicas: {}
          f:resources:
            f:limits:
              f:cpu: {}
              f:memory: {}
            f:requests:
              f:cpu: {}
              f:memory: {}
          f:retention: {}
          f:ruleSelector: {}
          f:rules:
            f:alert: {}
          f:securityContext:
            f:fsGroup: {}
            f:runAsNonRoot: {}
            f:runAsUser: {}
          f:serviceAccountName: {}
          f:serviceMonitorSelector: {}
          f:thanos:
            f:baseImage: {}
            f:resources: {}
            f:version: {}
          f:tsdb: {}
      manager: observability-operator
      operation: Apply
    - apiVersion: monitoring.rhobs/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          .: {}
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Reconciled"}:
              .: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:paused: {}
          f:replicas: {}
          f:shardStatuses:
            .: {}
            k:{"shardID":"0"}:
              .: {}
              f:availableReplicas: {}
              f:replicas: {}
              f:shardID: {}
              f:unavailableReplicas: {}
              f:updatedReplicas: {}
          f:unavailableReplicas: {}
          f:updatedReplicas: {}
      manager: PrometheusOperator
      operation: Update
      subresource: status

  3. Check the metadata.managedFields values, and observe that some fields in metadata and spec are managed by the MonitoringStack resource.
  4. Modify a field that is not controlled by the MonitoringStack resource:

    1. Change spec.enforcedSampleLimit, which is a field not set by the MonitoringStack resource. Create the file prom-spec-edited.yaml:

      prom-spec-edited.yaml

      apiVersion: monitoring.rhobs/v1
      kind: Prometheus
      metadata:
        name: sample-monitoring-stack
        namespace: coo-demo
      spec:
        enforcedSampleLimit: 1000

    2. Apply the YAML by running the following command:

      $ oc apply -f ./prom-spec-edited.yaml --server-side
      Note

      You must use the --server-side flag.

    3. Get the changed Prometheus object and note that there is one more section in managedFields which has spec.enforcedSampleLimit:

      $ oc get prometheus -n coo-demo

      Example output

      managedFields: 1
      - apiVersion: monitoring.rhobs/v1
        fieldsType: FieldsV1
        fieldsV1:
          f:metadata:
            f:labels:
              f:app.kubernetes.io/managed-by: {}
              f:app.kubernetes.io/name: {}
              f:app.kubernetes.io/part-of: {}
          f:spec:
            f:enforcedSampleLimit: {} 2
        manager: kubectl
        operation: Apply

      1
      managedFields
      2
      spec.enforcedSampleLimit
  5. Modify a field that is managed by the MonitoringStack resource:

    1. Change spec.LogLevel, which is a field managed by the MonitoringStack resource, using the following YAML configuration:

      # changing the logLevel from debug to info
      apiVersion: monitoring.rhobs/v1
      kind: Prometheus
      metadata:
        name: sample-monitoring-stack
        namespace: coo-demo
      spec:
        logLevel: info 1
      1
      spec.logLevel has been added
    2. Apply the YAML by running the following command:

      $ oc apply -f ./prom-spec-edited.yaml --server-side

      Example output

      error: Apply failed with 1 conflict: conflict with "observability-operator": .spec.logLevel
      Please review the fields above--they currently have other managers. Here
      are the ways you can resolve this warning:
      * If you intend to manage all of these fields, please re-run the apply
        command with the `--force-conflicts` flag.
      * If you do not intend to manage all of the fields, please edit your
        manifest to remove references to the fields that should keep their
        current managers.
      * You may co-own fields by updating your manifest to match the existing
        value; in this case, you'll become the manager if the other manager(s)
        stop managing the field (remove it from their configuration).
      See https://kubernetes.io/docs/reference/using-api/server-side-apply/#conflicts

    3. Notice that the field spec.logLevel cannot be changed using Server-Side Apply, because it is already managed by observability-operator.
    4. Use the --force-conflicts flag to force the change.

      $ oc apply -f ./prom-spec-edited.yaml --server-side --force-conflicts

      Example output

      prometheus.monitoring.rhobs/sample-monitoring-stack serverside-applied

      With --force-conflicts flag, the field can be forced to change, but since the same field is also managed by the MonitoringStack resource, the Observability Operator detects the change, and reverts it back to the value set by the MonitoringStack resource.

      Note

      Some Prometheus fields generated by the MonitoringStack resource are influenced by the fields in the MonitoringStack spec stanza, for example, logLevel. These can be changed by changing the MonitoringStack spec.

    5. To change the logLevel in the Prometheus object, apply the following YAML to change the MonitoringStack resource:

      apiVersion: monitoring.rhobs/v1alpha1
      kind: MonitoringStack
      metadata:
        name: sample-monitoring-stack
        labels:
          coo: example
      spec:
        logLevel: info
    6. To confirm that the change has taken place, query for the log level by running the following command:

      $ oc -n coo-demo get Prometheus.monitoring.rhobs -o=jsonpath='{.items[0].spec.logLevel}'

      Example output

      info

Note
  1. If a new version of an Operator generates a field that was previously generated and controlled by an actor, the value set by the actor will be overridden.

    For example, you are managing a field enforcedSampleLimit which is not generated by the MonitoringStack resource. If the Observability Operator is upgraded, and the new version of the Operator generates a value for enforcedSampleLimit, this will overide the value you have previously set.

  2. The Prometheus object generated by the MonitoringStack resource may contain some fields which are not explicitly set by the monitoring stack. These fields appear because they have default values.
Red Hat logoGithubRedditYoutubeTwitter

Learn

Try, buy, & sell

Communities

About Red Hat Documentation

We help Red Hat users innovate and achieve their goals with our products and services with content they can trust.

Making open source more inclusive

Red Hat is committed to replacing problematic language in our code, documentation, and web properties. For more details, see the Red Hat Blog.

About Red Hat

We deliver hardened solutions that make it easier for enterprises to work across platforms and environments, from the core datacenter to the network edge.

© 2024 Red Hat, Inc.