Merge pull request #619 from hashicorp/f-dc-empty

Alloc error message when no nodes are in DC

api/allocations.go

@@ -62,6 +62,7 @@ type Allocation struct {
 type AllocationMetric struct {
 	NodesEvaluated     int
 	NodesFiltered      int
+	NodesAvailable     map[string]int
 	ClassFiltered      map[string]int
 	ConstraintFiltered map[string]int
 	NodesExhausted     int

command/monitor.go

@@ -280,6 +280,14 @@ func dumpAllocStatus(ui cli.Ui, alloc *api.Allocation) {
 		ui.Output("  * No nodes were eligible for evaluation")
 	}
 
+	// Print a helpful message if the user has asked for a DC that has no
+	// available nodes.
+	for dc, available := range alloc.Metrics.NodesAvailable {
+		if available == 0 {
+			ui.Output(fmt.Sprintf("  * No nodes are available in datacenter %q", dc))
+		}
+	}
+
 	// Print filter info
 	for class, num := range alloc.Metrics.ClassFiltered {
 		ui.Output(fmt.Sprintf("  * Class %q filtered %d nodes", class, num))

nomad/structs/structs.go

@@ -1695,6 +1695,9 @@ type AllocMetric struct {
 	// NodesFiltered is the number of nodes filtered due to a constraint
 	NodesFiltered int
 
+	// NodesAvailable is the number of nodes available for evaluation per DC.
+	NodesAvailable map[string]int
+
 	// ClassFiltered is the number of nodes filtered by class
 	ClassFiltered map[string]int
 

scheduler/generic_sched.go

@@ -245,7 +245,7 @@ func (s *GenericScheduler) computeJobAllocs() error {
 // computePlacements computes placements for allocations
 func (s *GenericScheduler) computePlacements(place []allocTuple) error {
 	// Get the base nodes
-	nodes, err := readyNodesInDCs(s.state, s.job.Datacenters)
+	nodes, byDC, err := readyNodesInDCs(s.state, s.job.Datacenters)
 	if err != nil {
 		return err
 	}
@@ -279,6 +279,9 @@ func (s *GenericScheduler) computePlacements(place []allocTuple) error {
 			Metrics:   s.ctx.Metrics(),
 		}
 
+		// Store the available nodes by datacenter
+		s.ctx.Metrics().NodesAvailable = byDC
+
 		// Set fields based on if we found an allocation option
 		if option != nil {
 			// Generate the service ids for the tasks which this allocation is going
@@ -300,5 +303,6 @@ func (s *GenericScheduler) computePlacements(place []allocTuple) error {
 			failedTG[missing.TaskGroup] = alloc
 		}
 	}
+
 	return nil
 }

scheduler/generic_sched_test.go

@@ -123,6 +123,11 @@ func TestServiceSched_JobRegister_AllocFail(t *testing.T) {
 		t.Fatalf("bad: %#v", out[0].Metrics)
 	}
 
+	// Check the available nodes
+	if count, ok := out[0].Metrics.NodesAvailable["dc1"]; !ok || count != 0 {
+		t.Fatalf("bad: %#v", out[0].Metrics)
+	}
+
 	h.AssertEvalStatus(t, structs.EvalStatusComplete)
 }
 

scheduler/system_sched.go

@@ -25,12 +25,13 @@ type SystemScheduler struct {
 	state   State
 	planner Planner
 
-	eval  *structs.Evaluation
-	job   *structs.Job
-	plan  *structs.Plan
-	ctx   *EvalContext
-	stack *SystemStack
-	nodes []*structs.Node
+	eval      *structs.Evaluation
+	job       *structs.Job
+	plan      *structs.Plan
+	ctx       *EvalContext
+	stack     *SystemStack
+	nodes     []*structs.Node
+	nodesByDC map[string]int
 
 	limitReached bool
 	nextEval     *structs.Evaluation
@@ -86,7 +87,7 @@ func (s *SystemScheduler) process() (bool, error) {
 
 	// Get the ready nodes in the required datacenters
 	if s.job != nil {
-		s.nodes, err = readyNodesInDCs(s.state, s.job.Datacenters)
+		s.nodes, s.nodesByDC, err = readyNodesInDCs(s.state, s.job.Datacenters)
 		if err != nil {
 			return false, fmt.Errorf("failed to get ready nodes: %v", err)
 		}
@@ -219,7 +220,7 @@ func (s *SystemScheduler) computePlacements(place []allocTuple) error {
 			return fmt.Errorf("could not find node %q", missing.Alloc.NodeID)
 		}
 
-		// Update the set of placement ndoes
+		// Update the set of placement nodes
 		nodes[0] = node
 		s.stack.SetNodes(nodes)
 
@@ -246,6 +247,9 @@ func (s *SystemScheduler) computePlacements(place []allocTuple) error {
 			Metrics:   s.ctx.Metrics(),
 		}
 
+		// Store the available nodes by datacenter
+		s.ctx.Metrics().NodesAvailable = s.nodesByDC
+
 		// Set fields based on if we found an allocation option
 		if option != nil {
 			// Generate the service ids for the tasks that this allocation is going

scheduler/system_sched_test.go

@@ -59,6 +59,11 @@ func TestSystemSched_JobRegister(t *testing.T) {
 		t.Fatalf("bad: %#v", out)
 	}
 
+	// Check the available nodes
+	if count, ok := out[0].Metrics.NodesAvailable["dc1"]; !ok || count != 10 {
+		t.Fatalf("bad: %#v", out[0].Metrics)
+	}
+
 	h.AssertEvalStatus(t, structs.EvalStatusComplete)
 }
 

scheduler/util.go

@@ -172,19 +172,20 @@ func diffSystemAllocs(job *structs.Job, nodes []*structs.Node, taintedNodes map[
 	return result
 }
 
-// readyNodesInDCs returns all the ready nodes in the given datacenters
-func readyNodesInDCs(state State, dcs []string) ([]*structs.Node, error) {
+// readyNodesInDCs returns all the ready nodes in the given datacenters and a
+// mapping of each data center to the count of ready nodes.
+func readyNodesInDCs(state State, dcs []string) ([]*structs.Node, map[string]int, error) {
 	// Index the DCs
-	dcMap := make(map[string]struct{}, len(dcs))
+	dcMap := make(map[string]int, len(dcs))
 	for _, dc := range dcs {
-		dcMap[dc] = struct{}{}
+		dcMap[dc] = 0
 	}
 
 	// Scan the nodes
 	var out []*structs.Node
 	iter, err := state.Nodes()
 	if err != nil {
-		return nil, err
+		return nil, nil, err
 	}
 	for {
 		raw := iter.Next()
@@ -204,8 +205,9 @@ func readyNodesInDCs(state State, dcs []string) ([]*structs.Node, error) {
 			continue
 		}
 		out = append(out, node)
+		dcMap[node.Datacenter] += 1
 	}
-	return out, nil
+	return out, dcMap, nil
 }
 
 // retryMax is used to retry a callback until it returns success or

scheduler/util_test.go

@@ -204,7 +204,7 @@ func TestReadyNodesInDCs(t *testing.T) {
 	noErr(t, state.UpsertNode(1002, node3))
 	noErr(t, state.UpsertNode(1003, node4))
 
-	nodes, err := readyNodesInDCs(state, []string{"dc1", "dc2"})
+	nodes, dc, err := readyNodesInDCs(state, []string{"dc1", "dc2"})
 	if err != nil {
 		t.Fatalf("err: %v", err)
 	}
@@ -215,6 +215,12 @@ func TestReadyNodesInDCs(t *testing.T) {
 	if nodes[0].ID == node3.ID || nodes[1].ID == node3.ID {
 		t.Fatalf("Bad: %#v", nodes)
 	}
+	if count, ok := dc["dc1"]; !ok || count != 1 {
+		t.Fatalf("Bad: dc1 count %v", count)
+	}
+	if count, ok := dc["dc2"]; !ok || count != 1 {
+		t.Fatalf("Bad: dc2 count %v", count)
+	}
 }
 
 func TestRetryMax(t *testing.T) {