Fix the clusterstatus based on last operatorAction

Author: prafull01

.bazelrc

@@ -1,2 +1,2 @@
 build --workspace_status_command hack/build/print-workspace-status.sh
-test --test_output=errors --test_timeout=-1,-1,-1,2400
+test --test_output=all --test_timeout=-1,-1,-1,3600

pkg/actor/validate_version.go

@@ -184,11 +184,15 @@ func (v *versionChecker) Act(ctx context.Context, cluster *resource.Cluster, log
 			}
 			return errors.Wrapf(err, "failed to check the version of the crdb")
 		}
+
+		// Tail only last two lines as one line contains version info and one line has newline character. It is much faster.
+		podLines := int64(2)
 		podLogOpts := corev1.PodLogOptions{
 			Container: resource.JobContainerName,
+			TailLines: &podLines,
 		}
-		//get pod for the job we created
 
+		//get pod for the job we created
 		pods, err := v.clientset.CoreV1().Pods(job.Namespace).List(ctx, metav1.ListOptions{
 			LabelSelector: labels.Set(job.Spec.Selector.MatchLabels).AsSelector().String(),
 		})
@@ -231,11 +235,12 @@ func (v *versionChecker) Act(ctx context.Context, cluster *resource.Cluster, log
 		}
 		output := buf.String()
 
+		log.Info("Logs of version checker pod", "log", output)
 		// This is the value from Build Tag taken from the container
 		calVersion = strings.Replace(output, "\n", "", -1)
-		// if no image is retrieved we exit
+		// if no logs are found we will retry later to fetch the logs
 		if calVersion == "" {
-			err := PermanentErr{Err: errors.New("failed to check the version of the cluster")}
+			err := NotReadyErr{Err: errors.New("failed to check the version of the cluster")}
 			log.Error(err, "crdb version not found")
 			return err
 		}

pkg/clusterstatus/clusterstatus.go

@@ -31,25 +31,19 @@ func SetClusterStatusOnFirstReconcile(status *api.CrdbClusterStatus) {
 	status.ClusterStatus = api.ActionStatus(api.Starting).String()
 }
 
-// TODO: Do we need to check for all OperatorActions or for just the 0th element in SetClusterStatus func
-// nolint
+// SetClusterStatus will set the status of the cluster based on the status of the operator actions
+// operatorActions are not always appended, rather any particular action type is replaced if its status changes.
+// The status of the cluster is set to the status of the last operator action based on lastTransitionTime.
 func SetClusterStatus(status *api.CrdbClusterStatus) {
 	InitOperatorActionsIfNeeded(status, metav1.Now())
-	for _, a := range status.OperatorActions {
-		if a.Status == api.ActionStatus(api.Failed).String() {
-			status.ClusterStatus = api.ActionStatus(api.Failed).String()
-			return
+	for i := range status.OperatorActions {
+		if i == 0 || status.OperatorActions[i-1].LastTransitionTime.Before(&status.OperatorActions[i].LastTransitionTime) {
+			status.ClusterStatus = status.OperatorActions[i].Status
 		}
-
-		if a.Status == api.ActionStatus(api.Unknown).String() {
-			status.ClusterStatus = api.ActionStatus(api.Unknown).String()
-			return
-		}
-
-		status.ClusterStatus = api.ActionStatus(api.Finished).String()
-		return
 	}
-	status.ClusterStatus = api.ActionStatus(api.Finished).String()
+	if status.ClusterStatus == "" {
+		status.ClusterStatus = api.ActionStatus(api.Starting).String()
+	}
 }
 
 func InitOperatorActionsIfNeeded(status *api.CrdbClusterStatus, now metav1.Time) {