Use Service Placement Constraints in Enforcer

This patch updates the ConstraintEnforcer to use the placement
constraints from the current service spec rather than the placement
constraints from the current task spec because they may be outdated
in some cases (e.g., when the service was previously updated to
modify placement constrainst but the node to which the task was
scheduled still satisfies the constraints. If the node is updated
in a way which causes it to no longer satisfy the new constraints
then the task should be removed even if it still would satisfy the
original task constraints).

Docker-DCO-1.1-Signed-off-by: Josh Hawn <josh.hawn@docker.com> (github: jlhawn)

manager/orchestrator/constraintenforcer/constraint_enforcer.go

@@ -76,8 +76,22 @@ func (ce *ConstraintEnforcer) rejectNoncompliantTasks(node *api.Node) {
 		err   error
 	)
 
+	services := map[string]*api.Service{}
 	ce.store.View(func(tx store.ReadTx) {
 		tasks, err = store.FindTasks(tx, store.ByNodeID(node.ID))
+		if err != nil {
+			return
+		}
+
+		// Deduplicate service IDs using the services map. It's okay for the
+		// values to be nil for now, we will look them up from the store next.
+		for _, task := range tasks {
+			services[task.ServiceID] = nil
+		}
+
+		for serviceID := range services {
+			services[serviceID] = store.GetService(tx, serviceID)
+		}
 	})
 
 	if err != nil {
@@ -105,10 +119,44 @@ loop:
 			continue
 		}
 
-		// Ensure that the task still meets scheduling
-		// constraints.
-		if t.Spec.Placement != nil && len(t.Spec.Placement.Constraints) != 0 {
-			constraints, _ := constraint.Parse(t.Spec.Placement.Constraints)
+		// Ensure that the node still satisfies placement constraints.
+		// NOTE: If the task is associacted with a service then we must use the
+		// constraints from the current service spec rather than the
+		// constraints from the task spec because they may be outdated. This
+		// will happen if the service was previously updated in a way which
+		// only changes the placement constraints and the node matched the
+		// placement constraints both before and after that update. In the case
+		// of such updates, the tasks are not considered "dirty" and are not
+		// restarted but it will mean that the task spec's placement
+		// constraints are outdated. Consider this example:
+		// - A service is created with no constraints and a task is scheduled
+		//   to a node.
+		// - The node is updated to add a label, this doesn't affect the task
+		//   on that node because it has no constraints.
+		// - The service is updated to add a node label constraint which
+		//   matches the label which was just added to the node. The updater
+		//   does not shut down the task because the only the constraints have
+		//   changed and the node still matches the updated constraints.
+		// - The node is updated to remove the node label. The node no longer
+		//   satisfies the placement constraints of the service, so the task
+		//   should be shutdown. However, the task's spec still has the
+		//   original and outdated constraints (that are still satisfied by
+		//   the node). If we used those original constraints then the task
+		//   would incorrectly not be removed. This is why the constraints
+		//   from the service spec should be used instead.
+		var placement *api.Placement
+		if service := services[t.ServiceID]; service != nil {
+			// This task is associated with a service, so we use the service's
+			// current placement constraints.
+			placement = service.Spec.Task.Placement
+		} else {
+			// This task is not associated with a service (or the service no
+			// longer exists), so we use the placement constraints from the
+			// original task spec.
+			placement = t.Spec.Placement
+		}
+		if placement != nil && len(placement.Constraints) > 0 {
+			constraints, _ := constraint.Parse(placement.Constraints)
 			if !constraint.NodeMatches(constraints, node) {
 				removeTasks[t.ID] = t
 				continue

manager/orchestrator/constraintenforcer/constraint_enforcer_test.go

@@ -8,6 +8,7 @@ import (
 	"github.com/docker/swarmkit/manager/state"
 	"github.com/docker/swarmkit/manager/state/store"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
 func TestConstraintEnforcer(t *testing.T) {
@@ -168,3 +169,110 @@ func TestConstraintEnforcer(t *testing.T) {
 	assert.Equal(t, "id4", shutdown3.ID)
 	assert.Equal(t, api.TaskStateRejected, shutdown3.Status.State)
 }
+
+// TestOutdatedPlacementConstraints tests the following scenario: If a task is
+// associacted with a service then we must use the constraints from the current
+// service spec rather than the constraints from the task spec because they may
+// be outdated. This will happen if the service was previously updated in a way
+// which only changes the placement constraints and the node matched the
+// placement constraints both before and after that update. In the case of such
+// updates, the tasks are not considered "dirty" and are not restarted but it
+// will mean that the task spec's placement constraints are outdated. Consider
+// this example:
+// - A service is created with no constraints and a task is scheduled
+//   to a node.
+// - The node is updated to add a label, this doesn't affect the task
+//   on that node because it has no constraints.
+// - The service is updated to add a node label constraint which
+//   matches the label which was just added to the node. The updater
+//   does not shut down the task because the only the constraints have
+//   changed and the node still matches the updated constraints.
+// This test initializes a new in-memory store with the expected state from
+// above, starts a new constraint enforcer, and then updates the node to remove
+// the node label. Since the node no longer satisfies the placement constraints
+// of the service spec, the task should be shutdown despite the fact that the
+// task's own spec still has the original placement constraints.
+func TestOutdatedTaskPlacementConstraints(t *testing.T) {
+	node := &api.Node{
+		ID: "id0",
+		Spec: api.NodeSpec{
+			Annotations: api.Annotations{
+				Name: "node1",
+				Labels: map[string]string{
+					"foo": "bar",
+				},
+			},
+			Availability: api.NodeAvailabilityActive,
+		},
+		Status: api.NodeStatus{
+			State: api.NodeStatus_READY,
+		},
+		Role: api.NodeRoleWorker,
+	}
+
+	service := &api.Service{
+		ID: "id1",
+		Spec: api.ServiceSpec{
+			Annotations: api.Annotations{
+				Name: "service1",
+			},
+			Task: api.TaskSpec{
+				Placement: &api.Placement{
+					Constraints: []string{
+						"node.labels.foo == bar",
+					},
+				},
+			},
+		},
+	}
+
+	task := &api.Task{
+		ID: "id2",
+		Spec: api.TaskSpec{
+			Placement: nil, // Note: No placement constraints.
+		},
+		ServiceID: service.ID,
+		NodeID:    node.ID,
+		Status: api.TaskStatus{
+			State: api.TaskStateRunning,
+		},
+		DesiredState: api.TaskStateRunning,
+	}
+
+	s := store.NewMemoryStore(nil)
+	require.NotNil(t, s)
+	defer s.Close()
+
+	require.NoError(t, s.Update(func(tx store.Tx) error {
+		// Prepoulate node, service, and task.
+		for _, err := range []error{
+			store.CreateNode(tx, node),
+			store.CreateService(tx, service),
+			store.CreateTask(tx, task),
+		} {
+			if err != nil {
+				return err
+			}
+		}
+		return nil
+	}))
+
+	watch, cancel := state.Watch(s.WatchQueue(), api.EventUpdateTask{})
+	defer cancel()
+
+	constraintEnforcer := New(s)
+	defer constraintEnforcer.Stop()
+
+	go constraintEnforcer.Run()
+
+	// Update the node to remove the node label.
+	require.NoError(t, s.Update(func(tx store.Tx) error {
+		node = store.GetNode(tx, node.ID)
+		delete(node.Spec.Annotations.Labels, "foo")
+		return store.UpdateNode(tx, node)
+	}))
+
+	// The task should be rejected immediately.
+	task = testutils.WatchTaskUpdate(t, watch)
+	assert.Equal(t, api.TaskStateRejected, task.Status.State)
+}