Monitor your clusters

Gaining more insights into the state of the Kubernetes control plane was a highly requested feature from STACKIT Kubernetes Engine (SKE) customers. With Observability, STACKIT already provides a managed monitoring solution and we created an easy-to-use integration between our STACKIT services.

Connecting an Observability service with SKE

You need an existing Observability instance to use the monitoring.

As with all STACKIT Kubernetes Engine features, you have multiple options to activate and configure them: the Portal, the SKE API, the STACKIT CLI, or Terraform.

Portal

1. Navigate to the Kubernetes overview

2. Select an existing cluster or create a new one

Existing cluster

For an existing SKE cluster, navigate the new “Monitoring” section in the STACKIT Cloud Portal and select an Observability instance

New cluster

For new clusters, select an existing Observability instance during the creation process underneath the “Monitoring” section.

API

To configure the Observability integration via the API, you need to obtain the Observability instance ID. You can find this ID in the Observability dashboard in the API section, named Instance ID.

For both new and existing clusters, specify a cluster JSON object, and add the following to the extensions field:

{
  "extensions": {
    "observability": {
      "instanceId": "<observability instance id from portal>",
      "enabled": true
    }
  }
}

Note: The SKE API v2 uses the field instanceId. The deprecated API v1 only exposes extensions.argus (with argusInstanceId), which is replaced by extensions.observability. Do not use the argus extension for new setups.

For existing clusters, retrieve the current cluster configuration first, add only the extensions.observability section, and send the update. Do not remove or overwrite other extension fields.

CLI

1. Generate a cluster payload using the STACKIT CLI.

2. Add the extensions.observability section to the payload (see the API example above).

3. Update the cluster using stackit ske cluster update.

   stackit ske cluster update -p <PROJECT_ID> <CLUSTER_NAME> --payload @cluster-configuration.json

4. Check the reconcile status and the Grafana dashboard.

Terraform

extensions = {
  observability = {
    enabled     = true
    instance_id = var.observability_instance_id
  }
}

See the Terraform stackit_ske_cluster documentation for details.

The Observability extension only collects metrics from the kube-system namespace, with the exception of some cluster-wide metrics and node metrics. If you want to ingest metrics for your own application workloads, you have to set them up separately.

Recommendations on Observability instance sizing

The volume of metrics that the SKE service exposes is quite high, and also depends on the number of Pods, Nodes and Services. Therefore, we can give no one-size-fits-all recommendation for how to size your Observability instance. Nevertheless, you should always select at least a Basic Plan, even for small clusters with little workload. Larger clusters potentially require larger plans. If you have questions or need assistance sizing your Observability instance, don’t hesitate to contact our customer support.

Viewing Kubernetes metrics in Grafana

For your convenience, Observability provides a default Kubernetes dashboard for Grafana out of the box. You can find it in the STACKIT directory after signing in to your Grafana instance.

Further information

• Observability