multiregion: all regions start in running if no max_parallel

If `max_parallel` is not set, all regions should begin in a `running` state
rather than a `pending` state. Otherwise the first region is set to `running`
and then all the remaining regions once it enters `blocked. That behavior is
technically correct in that we have at most `max_parallel` regions running,
but definitely not what a user expects.

scheduler/reconcile.go

@@ -198,7 +198,6 @@ func (a *allocReconciler) Compute() *reconcileResults {
 	// Detect if the deployment is paused
 	if a.deployment != nil {
 		a.deploymentPaused = a.deployment.Status == structs.DeploymentStatusPaused
-		//||			a.deployment.Status == structs.DeploymentStatusPending
 		a.deploymentFailed = a.deployment.Status == structs.DeploymentStatusFailed
 	}
 	if a.deployment == nil {
@@ -555,9 +554,12 @@ func (a *allocReconciler) computeGroup(group string, all allocSet) bool {
 		// A previous group may have made the deployment already
 		if a.deployment == nil {
 			a.deployment = structs.NewDeployment(a.job)
-			// only the first region of a multiregion job starts in the
-			// running state
-			if a.job.IsMultiregion() && a.job.Region != a.job.Multiregion.Regions[0].Name {
+			// in a multiregion job, if max_parallel is set, only the first
+			// region starts in the running state
+			if a.job.IsMultiregion() &&
+				a.job.Multiregion.Strategy != nil &&
+				a.job.Multiregion.Strategy.MaxParallel != 0 &&
+				a.job.Region != a.job.Multiregion.Regions[0].Name {
 				a.deployment.Status = structs.DeploymentStatusPending
 				a.deployment.StatusDescription = structs.DeploymentStatusDescriptionPendingForPeer
 			}