Overriding resouce utilization measurements - Preparing to deploy hosted control planes | Hosted control planes

Overriding resource utilization measurements for a hosted cluster
Disabling the metric service monitoring

Hosted control planes is currently not available for OKD 4.20. The feature is planned to be released in the near future.

The set of baseline measurements for resource utilization can vary in each hosted cluster.

Overriding resource utilization measurements for a hosted cluster

You can override resource utilization measurements based on the type and pace of your cluster workload.

Procedure

Create the ConfigMap resource by running the following command:
```
$ oc create -f <your-config-map-file.yaml>
```
Replace <your-config-map-file.yaml> with the name of your YAML file that contains your hcp-sizing-baseline config map.

Create the hcp-sizing-baseline config map in the local-cluster namespace to specify the measurements you want to override. Your config map might resemble the following YAML file:

kind: ConfigMap
apiVersion: v1
metadata:
  name: hcp-sizing-baseline
  namespace: local-cluster
data:
  incrementalCPUUsagePer1KQPS: "9.0"
  memoryRequestPerHCP: "18"
  minimumQPSPerHCP: "50.0"

Delete the hypershift-addon-agent deployment to restart the hypershift-addon-agent pod by running the following command:
```
$ oc delete deployment hypershift-addon-agent \
  -n open-cluster-management-agent-addon
```

Verification

Observe the hypershift-addon-agent pod logs. Verify that the overridden measurements are updated in the config map by running the following command:

$ oc logs hypershift-addon-agent -n open-cluster-management-agent-addon

Your logs might resemble the following output:

Example output

2024-01-05T19:41:05.392Z	INFO	agent.agent-reconciler	agent/agent.go:793	setting cpuRequestPerHCP to 5
2024-01-05T19:41:05.392Z	INFO	agent.agent-reconciler	agent/agent.go:802	setting memoryRequestPerHCP to 18
2024-01-05T19:53:54.070Z	INFO	agent.agent-reconciler	agent/hcp_capacity_calculation.go:141	The worker nodes have 12.000000 vCPUs
2024-01-05T19:53:54.070Z	INFO	agent.agent-reconciler	agent/hcp_capacity_calculation.go:142	The worker nodes have 49.173369 GB memory

If the overridden measurements are not updated properly in the hcp-sizing-baseline config map, you might see the following error message in the hypershift-addon-agent pod logs:

Example error

2024-01-05T19:53:54.052Z	ERROR	agent.agent-reconciler	agent/agent.go:788	failed to get configmap from the hub. Setting the HCP sizing baseline with default values.	{"error": "configmaps \"hcp-sizing-baseline\" not found"}

Disabling the metric service monitoring

After you enable the hypershift-addon managed cluster add-on, metric service monitoring is configured by default so that OKD monitoring can gather metrics from hypershift-addon.

Procedure

You can disable metric service monitoring by completing the following steps:

Log in to your hub cluster by running the following command:
```
$ oc login
```
Edit the hypershift-addon-deploy-config add-on deployment configuration specification by running the following command:
```
$ oc edit addondeploymentconfig hypershift-addon-deploy-config \
  -n multicluster-engine
```

Add the disableMetrics=true customized variable to the specification, as shown in the following example:

apiVersion: addon.open-cluster-management.io/v1alpha1
kind: AddOnDeploymentConfig
metadata:
  name: hypershift-addon-deploy-config
  namespace: multicluster-engine
spec:
  customizedVariables:
  - name: hcMaxNumber
    value: "80"
  - name: hcThresholdNumber
    value: "60"
  - name: disableMetrics (1)
    value: "true"

1	The `disableMetrics=true` customized variable disables metric service monitoring for both new and existing `hypershift-addon` managed cluster add-ons.

Apply the changes to the configuration specification by running the following command:
```
$ oc apply -f <filename>.yaml
```