allocator: Less aggressive retry

Instead of retrying unallocated tasks, services, and networks every time
data changes in the store, limit these retries to every 5 minutes.

When a repeated attempt to allocate one of these objects fails, log it
at the debug log level, to reduce noise in the logs.

Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>

Author: Aaron Lehmann

manager/allocator/allocator_test.go

@@ -16,6 +16,11 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
+func init() {
+	// set artificially low retry interval for testing
+	retryInterval = 5 * time.Millisecond
+}
+
 func TestAllocator(t *testing.T) {
 	s := store.NewMemoryStore(nil)
 	assert.NotNil(t, s)

manager/allocator/network.go

@@ -26,7 +26,11 @@ const (
 	allocatedStatusMessage = "pending task scheduling"
 )
 
-var errNoChanges = errors.New("task unchanged")
+var (
+	errNoChanges = errors.New("task unchanged")
+
+	retryInterval = 5 * time.Minute
+)
 
 func newIngressNetwork() *api.Network {
 	return &api.Network{
@@ -57,19 +61,28 @@ type networkContext struct {
 	// the actual network allocation.
 	nwkAllocator *networkallocator.NetworkAllocator
 
-	// A table of unallocated tasks which will be revisited if any thing
+	// A set of tasks which are ready to be allocated as a batch. This is
+	// distinct from "unallocatedTasks" which are tasks that failed to
+	// allocate on the first try, being held for a future retry.
+	pendingTasks map[string]*api.Task
+
+	// A set of unallocated tasks which will be revisited if any thing
 	// changes in system state that might help task allocation.
 	unallocatedTasks map[string]*api.Task
 
-	// A table of unallocated services which will be revisited if
+	// A set of unallocated services which will be revisited if
 	// any thing changes in system state that might help service
 	// allocation.
 	unallocatedServices map[string]*api.Service
 
-	// A table of unallocated networks which will be revisited if
+	// A set of unallocated networks which will be revisited if
 	// any thing changes in system state that might help network
 	// allocation.
 	unallocatedNetworks map[string]*api.Network
+
+	// lastRetry is the last timestamp when unallocated
+	// tasks/services/networks were retried.
+	lastRetry time.Time
 }
 
 func (a *Allocator) doNetworkInit(ctx context.Context) (err error) {
@@ -80,10 +93,12 @@ func (a *Allocator) doNetworkInit(ctx context.Context) (err error) {
 
 	nc := &networkContext{
 		nwkAllocator:        na,
+		pendingTasks:        make(map[string]*api.Task),
 		unallocatedTasks:    make(map[string]*api.Task),
 		unallocatedServices: make(map[string]*api.Service),
 		unallocatedNetworks: make(map[string]*api.Network),
 		ingressNetwork:      newIngressNetwork(),
+		lastRetry:           time.Now(),
 	}
 	a.netCtx = nc
 	defer func() {
@@ -401,12 +416,22 @@ func (a *Allocator) doNetworkAlloc(ctx context.Context, ev events.Event) {
 	case state.EventCreateNode, state.EventUpdateNode, state.EventDeleteNode:
 		a.doNodeAlloc(ctx, ev)
 	case state.EventCreateTask, state.EventUpdateTask, state.EventDeleteTask:
-		a.doTaskAlloc(ctx, ev)
+		a.doTaskAlloc(ctx, ev, nc.pendingTasks)
 	case state.EventCommit:
-		a.procUnallocatedNetworks(ctx)
-		a.procUnallocatedServices(ctx)
-		a.procUnallocatedTasksNetwork(ctx)
-		return
+		a.procTasksNetwork(ctx, nc.pendingTasks, false)
+
+		if time.Since(nc.lastRetry) > retryInterval {
+			a.procUnallocatedNetworks(ctx)
+			a.procUnallocatedServices(ctx)
+			a.procTasksNetwork(ctx, nc.unallocatedTasks, true)
+			nc.lastRetry = time.Now()
+		}
+
+		// Any left over tasks are moved to the unallocated set
+		for _, t := range nc.pendingTasks {
+			nc.unallocatedTasks[t.ID] = t
+		}
+		nc.pendingTasks = make(map[string]*api.Task)
 	}
 }
 
@@ -551,7 +576,7 @@ func (a *Allocator) taskCreateNetworkAttachments(t *api.Task, s *api.Service) {
 	taskUpdateNetworks(t, networks)
 }
 
-func (a *Allocator) doTaskAlloc(ctx context.Context, ev events.Event) {
+func (a *Allocator) doTaskAlloc(ctx context.Context, ev events.Event, toAllocate map[string]*api.Task) {
 	var (
 		isDelete bool
 		t        *api.Task
@@ -579,14 +604,16 @@ func (a *Allocator) doTaskAlloc(ctx context.Context, ev events.Event) {
 			}
 		}
 
-		// Cleanup any task references that might exist in unallocatedTasks
+		// Cleanup any task references that might exist
+		delete(toAllocate, t.ID)
 		delete(nc.unallocatedTasks, t.ID)
 		return
 	}
 
 	// If we are already in allocated state, there is
 	// absolutely nothing else to do.
 	if t.Status.State >= api.TaskStatePending {
+		delete(toAllocate, t.ID)
 		delete(nc.unallocatedTasks, t.ID)
 		return
 	}
@@ -616,7 +643,7 @@ func (a *Allocator) doTaskAlloc(ctx context.Context, ev events.Event) {
 	// based on service spec.
 	a.taskCreateNetworkAttachments(t, s)
 
-	nc.unallocatedTasks[t.ID] = t
+	toAllocate[t.ID] = t
 }
 
 func (a *Allocator) allocateNode(ctx context.Context, node *api.Node) error {
@@ -948,15 +975,18 @@ func (a *Allocator) procUnallocatedServices(ctx context.Context) {
 	}
 }
 
-func (a *Allocator) procUnallocatedTasksNetwork(ctx context.Context) {
-	nc := a.netCtx
-	allocatedTasks := make([]*api.Task, 0, len(nc.unallocatedTasks))
+func (a *Allocator) procTasksNetwork(ctx context.Context, toAllocate map[string]*api.Task, quiet bool) {
+	allocatedTasks := make([]*api.Task, 0, len(toAllocate))
 
-	for _, t := range nc.unallocatedTasks {
+	for _, t := range toAllocate {
 		if err := a.allocateTask(ctx, t); err == nil {
 			allocatedTasks = append(allocatedTasks, t)
 		} else if err != errNoChanges {
-			log.G(ctx).WithError(err).Error("task allocation failure")
+			if quiet {
+				log.G(ctx).WithError(err).Debug("task allocation failure")
+			} else {
+				log.G(ctx).WithError(err).Error("task allocation failure")
+			}
 		}
 	}
 
@@ -978,11 +1008,11 @@ func (a *Allocator) procUnallocatedTasksNetwork(ctx context.Context) {
 	})
 
 	if err != nil {
-		log.G(ctx).WithError(err).Error("failed a store batch operation while processing unallocated tasks")
+		log.G(ctx).WithError(err).Error("failed a store batch operation while processing tasks")
 	}
 
 	for _, t := range allocatedTasks[:committed] {
-		delete(nc.unallocatedTasks, t.ID)
+		delete(toAllocate, t.ID)
 	}
 }