Skip to content

Latest commit

 

History

History
141 lines (108 loc) · 4.95 KB

CONTRIBUTING.md

File metadata and controls

141 lines (108 loc) · 4.95 KB
Raw

Contribute

Bootstrap Multiple Clusters Using the Log Storage Toolbox

In the Log Storage Toolbox project on GitLab (behind VPN), we have a set of scripts that allow us to provision OCP clusters in multiple cloud providers. The README.md under log-storage-toolbox/manifests/ocp contains detailed instructions on how to use the script. NOTE: Be sure to follow the instructions in the base section as they are prerequisites.

For development on the multicluster-observability-addon, most of the time it's handy to provision two clusters, one hub and one spoke.

# Download openshift-install client
./scripts/ocp-download-release.sh 4.14.7
# Prepare bootstrap resources
./scripts/ocp-install.sh aws eu-central-1 4.14.7
# Launch cluster (takes 40min +-)
openshift-install-linux-4.14.7 create cluster --dir ./output/jmarcalaws24011261
./scripts/ocp-install.sh aws eu-central-1 4.14.7
openshift-install-linux-4.14.7 create cluster --dir ./output/jmarcalaws24011221

Both clusters are supported by 6 nodes m6a.4xlarge (3 masters, 3 workers). It's indifferent which cluster you pick to be the hub and the spoke. When you are done with developing, be sure to destroy them.

openshift-install-linux-4.14.7 destroy cluster --dir ./output/jmarcalaws24011261
openshift-install-linux-4.14.7 destroy cluster --dir ./output/jmarcalaws24011221

Bootstrapping ACM and Linking Clusters

All steps are meant to be run on the hub cluster except when explicitly stated.

  1. Use the OpenShift Installer to create and set up two OCP clusters.
  2. Install the Advanced Cluster Management for Kubernetes operator.
  3. Create a MultiClusterHub resource using the web console.
  4. Import each spoke cluster to RHACM via the web console (top left, to the right of the RH OpenShift logo), using the commands option by running the commands on each spoke cluster.

Development Cycle for Smoke Testing

Note: the addon has a dependency on cert-manager operator, which should be installed on the hub cluster

When working on the addon, it's nice to be able to test things quickly; to do this, you can:

export REGISTRY_BASE=quay.io/YOUR_QUAY_ID
# Builds and pushes the addon images
make oci 
# Deploys the CRDs necessary, the addon using your built image
make addon-deploy

Then every time you want to test a new version, you can just:

make oci
# Delete the mcoa pod which will make the Deployment pull the new image
oc -n open-cluster-management delete pod -l app=multicluster-observability-addon-manager

Disabeling specific signals

The addon supports disabling signals using the resource AddOnDeploymentConfig. For instance, to disable the logging signal create the following resource on the hub cluster:

apiVersion: addon.open-cluster-management.io/v1alpha1
kind: AddOnDeploymentConfig
metadata:
  name: multicluster-observability-addon
  namespace: open-cluster-management
spec:
  customizedVariables:
    - name: loggingDisabled
      value: "true"

Supported keys are metricsDisabled, loggingDisabled and tracingDisabled

Install the addon on a Spoke Cluster

To actually install the addon on a spoke cluster, you need to:

  1. Have the addon manager running on the hub cluster.
  2. Create the necessary Kubernetes resources in the namespace of the spoke cluster that will be used by the addon to generate the ManifestWorks, e.g., secrets, configmaps.
  3. Create the ManagedClusterAddon resource in the namespace of the spoke cluster.
apiVersion: addon.open-cluster-management.io/v1alpha1
kind: ManagedClusterAddOn
metadata:
  name: multicluster-observability-addon
  namespace: spoke-1
spec:
  installNamespace: open-cluster-management-agent-addon
  configs:
  - resource: configmaps
    name: spoke-1
    namespace: spoke-1
  - resource: secrets
    name: spoke-1
    namespace: spoke-1
  1. Once a ManagedClusterAddon is reconciled successfuly by the addon we can look for the ManifestWorks
oc -n spoke-1 get manifestworks addon-multicluster-observability-addon-deploy-0

Configuring Metrics

Currently the addon doesn't support any configuration, so no configuration is needed at the ManagedClusterAddOn level. However, the addon has a dependency with MCO. Nowadays the addon supports the collection of metrics from the spoke clusters. These metrics are sent to an MCO instance running in the Hub.

Configuring Logs

Currently the addon supports configuration to send logs either to:

  • CloudWatch: requires the auth configmap to be specified
  • Loki: requires the auth configmap, the url configmap and optionally the inject ca configmap

Configuring Traces

Currently the addon supports configuration to send traces to:

  • OpenTelemetryCollector: requires the auth configmap, the url configmap and optionally the inject ca configmap