fix: serialize all workflow jobs

cd/manager/jobmanager/jobManager.go

@@ -261,7 +261,8 @@ func (m *JobManager) processForceDeployJobs(dequeuedJobs []job.JobState) bool {
 }
 
 func (m *JobManager) processDeployJobs(dequeuedJobs []job.JobState) bool {
-	// Check if there are any (non-anchor) jobs in progress
+	// Check if there are any non-anchor jobs in progress. Deployments can run in parallel with anchor jobs but not with
+	// any other jobs.
 	activeNonAnchorJobs := m.getActiveNonAnchorJobs()
 	if len(activeNonAnchorJobs) == 0 {
 		// We know the first job is a deploy, so pick out the component for that job, collapse as many back-to-back jobs
@@ -355,7 +356,7 @@ func (m *JobManager) processTestJobs(dequeuedJobs []job.JobState) bool {
 	if len(m.getActiveDeploys()) == 0 {
 		// - Collapse all smoke tests between deployments into a single run
 		// - Collapse all E2E tests between deployments into a single run
-		dequeuedTests := make(map[job.JobType]job.JobState, 0)
+		dequeuedTests := make(map[job.JobType]job.JobState)
 		for _, dequeuedJob := range dequeuedJobs {
 			// Break out of the loop as soon as we find a deploy job so that we don't collapse test jobs across deploys.
 			if dequeuedJob.Type == job.JobType_Deploy {
@@ -382,39 +383,17 @@ func (m *JobManager) processTestJobs(dequeuedJobs []job.JobState) bool {
 }
 
 func (m *JobManager) processWorkflowJobs(dequeuedJobs []job.JobState) bool {
-	// Check if there are any deploy jobs in progress
-	if len(m.getActiveDeploys()) == 0 {
-		dequeuedWorkflows := make([]job.JobState, 0)
-		// Do not collapse back-to-back workflow jobs because they could be pointing to different actual workflows
+	// Check if there are any non-anchor jobs in progress. Workflows can run in parallel with anchor jobs but not with
+	// any other jobs.
+	if len(m.getActiveNonAnchorJobs()) == 0 {
 		for _, dequeuedJob := range dequeuedJobs {
-			// Break out of the loop as soon as we find a deploy job so that we don't collapse workflow jobs across
-			// deploys.
-			if dequeuedJob.Type == job.JobType_Deploy {
-				break
-			} else if dequeuedJob.Type == job.JobType_Workflow {
-				if workflow, err := job.CreateWorkflowJob(dequeuedJob); err != nil {
-					if err = m.updateJobStage(dequeuedJob, job.JobStage_Failed, err); err != nil {
-						log.Printf("processWorkflowJobs: job update failed: %v, %s", err, manager.PrintJob(dequeuedJob))
-					}
-				} else if workflow.IsType(job.WorkflowJobLabel_Deploy) {
-					// If we haven't accumulated any workflow jobs yet, this is the first job in the list. If there were
-					// no other active jobs, queue it and break from the loop because we can't run any other jobs in
-					// parallel with deployments.
-					if (len(dequeuedWorkflows) == 0) && (len(m.getActiveNonAnchorJobs()) == 0) {
-						dequeuedWorkflows = append(dequeuedWorkflows, dequeuedJob)
-					}
-					// If there were some accumulated jobs by this point, we know they must be test jobs. We'll still
-					// break out of the loop here so that we don't run test jobs across deploy jobs.
-					break
-				} else {
-					dequeuedWorkflows = append(dequeuedWorkflows, dequeuedJob)
-				}
+			if dequeuedJob.Type == job.JobType_Workflow {
+				m.advanceJob(dequeuedJob)
+				return true
 			}
 		}
-		m.advanceJobs(dequeuedWorkflows)
-		return len(dequeuedWorkflows) > 0
 	} else {
-		log.Printf("processWorkflowJobs: deployment in progress")
+		log.Printf("processWorkflowJobs: other jobs in progress")
 	}
 	return false
 }