DCAE platform components (inventory, deployment handler, policy handler, etc.) are deployed and undeployed using Helm. DCAE service components--data collectors and data analytics modules--are deployed using Cloudify, with the DCAE k8s plugin. When DCAE is undeployed, Helm has no way to undeploy the service components. The artifacts in this directory build a Docker image that can be run as a Kubernetes Job, using a Helm pre-delete hook. The image runs a script that deletes the Kubernetes Services and Kubernetes Deployments (and all of the ReplicaSets and Pods created as children of the Deployments) that were created by the k8s plugin.
The script relies on the fact that Services and Deployments created by the k8s plugin have a unique label ("cfydeployment"). The script finds Services and Deployments with that label and deletes them.
The image is intended to be run as Kubernetes Job in a Helm pre-delete hook associated with the OOM dcaegen2 charts. A Helm template in the OOM dcaegen2 tree defines the Job. The Job will start a container. The container will execute the dcae-cleanup.sh script and then exit. The intent is that using a helm undeploy command will automatically delete all of the DCAE service components, so that no additional cleanup is needed.
The container can be run manually using the kubectl run command. For example:
kubectl -n onap run --restart='OnFailure' --image dcae-cleanup:0.0.0 cleanup
The --restart='OnFailure' parameter causes kubectl to create a Job.