Is it possible to deploy a clean up job after dependent resources are deleted?

[raminqaf]

Hello,
I am writing an Operator that the primary resource manages deployment, service, configMap, etc. The resource that I am creating needs a graceful shutdown. This can be achieved through a Job (i.e., clean-up job) resource and after all the secondary resources (deployment, service, configMap, etc.) are deleted. What should I do in my cleanup method to achieve this?

Can this be achieved by calling the noFinalizerRemoval() and then processing with the deployment of the clean-up job and then calling defaultDelete()?

    @Override
    public DeleteControl cleanup(final MyComplexApp resource, final Context<MyComplexApp> context) {
        DeleteControl.noFinalizerRemoval();
        
        // Deploy clean-up job (and wait to finish?) ...
        
       return DeleteControl.defaultDelete();
    }

[metacosm]

I guess it depends if you care about the result of the Job. Also, are you using the dependent resource feature?

[raminqaf]

@metacosm Yeah, I am using the dependent resource feature:

@ControllerConfiguration(
        dependents = {
                @Dependent(type = DeploymentDependentResource.class),
                //...
        })
public class MyComplexAppReconciler implements Reconciler< MyComplexApp >, ErrorStatusHandler< MyComplexApp >,
        Cleaner< MyComplexApp > {
        // ...
}

However, I am unsure if the Job should be defined as a dependent resource. For this example, the cleanup job responsibility is to delete a table in the database.

[csviri]

However, I am unsure if the Job should be defined as a dependent resource. For this example, the cleanup job responsibility is to delete a table in the database.

Porbably not,
what you mention here should be the way to go:

Can this be achieved by calling the noFinalizerRemoval() and then processing with the deployment of the clean-up job and then calling defaultDelete()?

[raminqaf]

@csviri I'll give it a shot and write an update, on how it went! Thanks!

[raminqaf]

@metacosm & @csviri I have updated the cleanup method, and now it looks like this. For simplicity, I am loading the cleanup job Yaml file staticly.

    @Override
    public DeleteControl cleanup(final ComplexApp complexApp, final Context<ComplexApp> context) {
        final DeleteControl deleteControl = DeleteControl.noFinalizerRemoval();

        final Job job = loadYaml(Job.class, Utils.class, "cleanup-job.yaml");
        final KubernetesClient client = context.getClient();
        final String namespace = complexApp.getMetadata().getNamespace();

        final Job jobInCluster = client.resource(job).inNamespace(namespace).get();

        if (jobInCluster == null) {
            client.resource(job)
                    .inNamespace(namespace)
                    .create();
            client.resource(job)
                    .waitUntilReady(10, TimeUnit.SECONDS);
        } else {
            if(jobInCluster.getStatus() != null && jobInCluster.getStatus().getSucceeded() == 1) {
                // Maybe even more checks and conditions to be checked here...
                return DeleteControl.defaultDelete();
            }
        }
        return deleteControl.rescheduleAfter(10, TimeUnit.SECONDS);
   }

I hoped the cleanup function would be called again after the rescheduled is returned. Am I missing something here?

[csviri]

I hoped the cleanup function would be called again after the rescheduled is returned. Am I missing something here?

No, that should be the case.

This is the integration test for it: https://github.com/java-operator-sdk/java-operator-sdk/blob/32586bc277301e4b08b8d04ed6e34e34b9291596/operator-framework/src/test/java/io/javaoperatorsdk/operator/CleanerForReconcilerIT.java#L46-L46

[raminqaf]

@csviri Thanks for the link! It helped me to understand the cleanup better. I want to delete all the dependent resources. After a successful deletion of the DRs, I want then check with a client API that a condition is set to deploy the clean-up job. And lastly, delete the Primary Resouce. Do you think this is also achievable?

    @Override
    public DeleteControl cleanup(final ComplexApp complexApp, final Context<ComplexApp> context) {
        // 1. Delete all the DRs
        
        // 2. If all the DRs are cleared
        
        // 3. Check with a client API if some conditions are set. If not reschedule.
        
        // 4. Deploy Clean up Job
        
        // 5. Call DeleteControl.defaultDelete();
   }

[csviri]

Absolutely, if you use managed DRs, (1) this should be done for you. Check also the current state of DRs either from context of simple the cache.

Note that the whole thing can be async.

[raminqaf]

@csviri Thanks! Could you please point me to some sample code or documentation regarding the state of the DRs?

[csviri]

@raminqaf You can check either just the current state from context. But though the context you can access this: https://github.com/java-operator-sdk/java-operator-sdk/blob/a6a57afd0079f4b64fed42292ef1298e3d7d5187/operator-framework-core/src/main/java/io/javaoperatorsdk/operator/api/reconciler/dependent/managed/ManagedDependentResourceContext.java#L55-L57

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days.

[github-actions]

This issue is stale because it has been open 60 days with no activity. Remove stale label or comment or this will be closed in 14 days.