Skip to content

Install with Helm (OCI)

The operator is published as two Helm charts in the GitHub Container Registry:

Chart Contents
oci://ghcr.io/gradata-systems/helm-charts/stroom-operator-crds The four CRDs.
oci://ghcr.io/gradata-systems/helm-charts/stroom-operator The controller Deployment, RBAC and ServiceAccount.

The CRDs are a separate chart so that uninstalling or reinstalling the controller never deletes them. Deleting a CRD deletes every resource of that kind in the cluster, including every StroomCluster.

Install

Install the CRDs first, then the operator:

helm install stroom-operator-crds \
  oci://ghcr.io/gradata-systems/helm-charts/stroom-operator-crds \
  --namespace stroom-operator-system --create-namespace

helm install stroom-operator \
  oci://ghcr.io/gradata-systems/helm-charts/stroom-operator \
  --namespace stroom-operator-system

Without --version, Helm installs the latest chart. To install a specific version, which is advisable in production, add --version <chart version> to both commands and use the same version for each. To find the latest version and the operator version it deploys (appVersion):

helm show chart oci://ghcr.io/gradata-systems/helm-charts/stroom-operator

Check the installation

kubectl -n stroom-operator-system rollout status deployment/stroom-operator-controller-manager
kubectl get crds | grep stroom.gchq.github.io

You should see four CRDs: stroomclusters, databaseservers, databasebackups and stroomtaskautoscalers.

To follow the operator's log:

kubectl -n stroom-operator-system logs deployment/stroom-operator-controller-manager -f

Chart values

The stroom-operator chart accepts these values:

Value Default Description
registry "" Private registry to pull the operator image from, in place of the registry in image.repository. The image becomes <registry>/gradata-systems/stroom-k8s-operator:<tag>.
image.repository ghcr.io/gradata-systems/stroom-k8s-operator Operator image repository.
image.tag Chart appVersion Operator image tag.
image.pullPolicy IfNotPresent Image pull policy.
resources Requests cpu: 500m, memory: 512Mi; limit memory: 1Gi Controller resources.
securityContext Non-root (UID/GID 65532), RuntimeDefault seccomp Pod security context.
nodeSelector, tolerations, affinity Empty Controller scheduling.

The stroom-operator-crds chart has no values.

For example, to give the controller more memory:

helm upgrade stroom-operator \
  oci://ghcr.io/gradata-systems/helm-charts/stroom-operator \
  --namespace stroom-operator-system \
  --reuse-values --set resources.limits.memory=2Gi

Controller memory

The StroomTaskAutoscaler keeps up to 60 minutes of CPU samples in memory for every processing node it manages. In large deployments, raise the controller's memory limit if it is OOMKilled.

Upgrade and removal

See Upgrading and Stopping and removal.