[Bugfix] Prevent deployment removal in case of invalid K8S API response (#1063)

Author: ajanikow

CHANGELOG.md

@@ -10,6 +10,7 @@
 - (Feature) Add 'crd install' subcommand
 - (Bugfix) Fix `internal` metrics mode
 - (Bugfix) Create agency dump if auth is disabled
+- (Bugfix) Prevent deployment removal in case of invalid K8S API response
 
 ## [1.2.14](https://github.com/arangodb/kube-arangodb/tree/1.2.14) (2022-07-14)
 - (Feature) Add ArangoSync TLS based rotation

pkg/deployment/deployment.go

@@ -293,9 +293,9 @@ func (d *Deployment) Update(apiObject *api.ArangoDeployment) {
 	})
 }
 
-// Delete the deployment.
+// Stop the deployment.
 // Called when the deployment was deleted by the user.
-func (d *Deployment) Delete() {
+func (d *Deployment) Stop() {
 	d.log.Info("deployment is deleted by user")
 	if atomic.CompareAndSwapInt32(&d.stopped, 0, 1) {
 		close(d.stopCh)
@@ -359,18 +359,6 @@ func (d *Deployment) run() {
 	for {
 		select {
 		case <-d.stopCh:
-			err := d.acs.CurrentClusterCache().Refresh(context.Background())
-			if err != nil {
-				log.Err(err).Error("Unable to get resources")
-			}
-			// Remove finalizers from created resources
-			log.Info("Deployment removed, removing finalizers to prevent orphaned resources")
-			if _, err := d.removePodFinalizers(context.TODO(), d.GetCachedStatus()); err != nil {
-				log.Err(err).Warn("Failed to remove Pod finalizers")
-			}
-			if _, err := d.removePVCFinalizers(context.TODO(), d.GetCachedStatus()); err != nil {
-				log.Err(err).Warn("Failed to remove PVC finalizers")
-			}
 			// We're being stopped.
 			return
 

pkg/deployment/deployment_inspector.go

@@ -74,7 +74,7 @@ func (d *Deployment) inspectDeployment(lastInterval util.Interval) util.Interval
 	if k8sutil.IsNotFound(err) {
 		// Deployment is gone
 		d.log.Info("Deployment is gone")
-		d.Delete()
+		d.Stop()
 		return nextInterval
 	} else if updated != nil && updated.GetDeletionTimestamp() != nil {
 		// Deployment is marked for deletion

pkg/operator/operator_deployment.go

@@ -185,7 +185,7 @@ func (o *Operator) handleDeploymentEvent(event *Event) error {
 		if !ok {
 			return errors.WithStack(errors.Newf("unsafe state. deployment (%s) was never created but we received event (%s)", apiObject.Name, event.Type))
 		}
-		depl.Delete()
+		depl.Stop()
 		delete(o.deployments, apiObject.Name)
 		deploymentsDeleted.Inc()
 		deploymentsCurrent.Set(float64(len(o.deployments)))