Expose a pending pods summary in scheudler's dummper output

This commit is contained in:
Wei Huang 2022-08-05 18:30:09 -07:00
parent 64ed914545
commit 7df9bfcfef
No known key found for this signature in database
GPG Key ID: 17AFE05D01EA77B2
7 changed files with 30 additions and 17 deletions

View File

@ -54,7 +54,7 @@ func (c *CacheComparer) Compare() error {
dump := c.Cache.Dump()
pendingPods := c.PodQueue.PendingPods()
pendingPods, _ := c.PodQueue.PendingPods()
if missed, redundant := c.CompareNodes(nodes, dump.Nodes); len(missed)+len(redundant) != 0 {
klog.InfoS("Cache mismatch", "missedNodes", missed, "redundantNodes", redundant)

View File

@ -54,12 +54,12 @@ func (d *CacheDumper) dumpNodes() {
// dumpSchedulingQueue writes pods in the scheduling queue to the scheduler logs.
func (d *CacheDumper) dumpSchedulingQueue() {
pendingPods := d.podQueue.PendingPods()
pendingPods, s := d.podQueue.PendingPods()
var podData strings.Builder
for _, p := range pendingPods {
podData.WriteString(printPod(p))
}
klog.InfoS("Dump of scheduling queue", "pods", podData.String())
klog.InfoS("Dump of scheduling queue", "summary", s, "pods", podData.String())
}
// printNodeInfo writes parts of NodeInfo to a string.

View File

@ -102,7 +102,7 @@ type SchedulingQueue interface {
MoveAllToActiveOrBackoffQueue(event framework.ClusterEvent, preCheck PreEnqueueCheck)
AssignedPodAdded(pod *v1.Pod)
AssignedPodUpdated(pod *v1.Pod)
PendingPods() []*v1.Pod
PendingPods() ([]*v1.Pod, string)
// Close closes the SchedulingQueue so that the goroutine which is
// waiting to pop items can exit gracefully.
Close()
@ -678,9 +678,12 @@ func (p *PriorityQueue) getUnschedulablePodsWithMatchingAffinityTerm(pod *v1.Pod
return podsToMove
}
// PendingPods returns all the pending pods in the queue. This function is
// used for debugging purposes in the scheduler cache dumper and comparer.
func (p *PriorityQueue) PendingPods() []*v1.Pod {
var pendingPodsSummary = "activeQ:%v; backoffQ:%v; unschedulablePods:%v"
// PendingPods returns all the pending pods in the queue; accompanied by a debugging string
// recording showing the number of pods in each queue respectively.
// This function is used for debugging purposes in the scheduler cache dumper and comparer.
func (p *PriorityQueue) PendingPods() ([]*v1.Pod, string) {
p.lock.RLock()
defer p.lock.RUnlock()
var result []*v1.Pod
@ -693,7 +696,7 @@ func (p *PriorityQueue) PendingPods() []*v1.Pod {
for _, pInfo := range p.unschedulablePods.podInfoMap {
result = append(result, pInfo.Pod)
}
return result
return result, fmt.Sprintf(pendingPodsSummary, p.activeQ.Len(), p.podBackoffQ.Len(), len(p.unschedulablePods.podInfoMap))
}
// Close closes the priority queue.

View File

@ -745,13 +745,21 @@ func TestPriorityQueue_PendingPods(t *testing.T) {
q.AddUnschedulableIfNotPresent(q.newQueuedPodInfo(highPriorityPodInfo.Pod), q.SchedulingCycle())
expectedSet := makeSet([]*v1.Pod{medPriorityPodInfo.Pod, unschedulablePodInfo.Pod, highPriorityPodInfo.Pod})
if !reflect.DeepEqual(expectedSet, makeSet(q.PendingPods())) {
gotPods, gotSummary := q.PendingPods()
if !reflect.DeepEqual(expectedSet, makeSet(gotPods)) {
t.Error("Unexpected list of pending Pods.")
}
if wantSummary := fmt.Sprintf(pendingPodsSummary, 1, 0, 2); wantSummary != gotSummary {
t.Errorf("Unexpected pending pods summary: want %v, but got %v.", wantSummary, gotSummary)
}
// Move all to active queue. We should still see the same set of pods.
q.MoveAllToActiveOrBackoffQueue(TestEvent, nil)
if !reflect.DeepEqual(expectedSet, makeSet(q.PendingPods())) {
t.Error("Unexpected list of pending Pods...")
gotPods, gotSummary = q.PendingPods()
if !reflect.DeepEqual(expectedSet, makeSet(gotPods)) {
t.Error("Unexpected list of pending Pods.")
}
if wantSummary := fmt.Sprintf(pendingPodsSummary, 1, 2, 0); wantSummary != gotSummary {
t.Errorf("Unexpected pending pods summary: want %v, but got %v.", wantSummary, gotSummary)
}
}

View File

@ -418,7 +418,7 @@ func TestFailureHandler_PodAlreadyBound(t *testing.T) {
// getPodFromPriorityQueue is the function used in the TestDefaultErrorFunc test to get
// the specific pod from the given priority queue. It returns the found pod in the priority queue.
func getPodFromPriorityQueue(queue *internalqueue.PriorityQueue, pod *v1.Pod) *v1.Pod {
podList := queue.PendingPods()
podList, _ := queue.PendingPods()
if len(podList) == 0 {
return nil
}

View File

@ -1713,9 +1713,9 @@ func TestUnschedulablePodBecomesSchedulable(t *testing.T) {
t.Errorf("Pod %v was not scheduled: %v", pod.Name, err)
}
// Make sure pending queue is empty.
pendingPods := len(testCtx.Scheduler.SchedulingQueue.PendingPods())
if pendingPods != 0 {
t.Errorf("pending pods queue is not empty, size is: %d", pendingPods)
pendingPods, s := testCtx.Scheduler.SchedulingQueue.PendingPods()
if len(pendingPods) != 0 {
t.Errorf("pending pods queue is not empty, size is: %d, summary is: %s", len(pendingPods), s)
}
})
}

View File

@ -88,7 +88,8 @@ func TestCoreResourceEnqueue(t *testing.T) {
// Wait for the three pods to be present in the scheduling queue.
if err := wait.Poll(time.Millisecond*200, wait.ForeverTestTimeout, func() (bool, error) {
return len(testCtx.Scheduler.SchedulingQueue.PendingPods()) == 3, nil
pendingPods, _ := testCtx.Scheduler.SchedulingQueue.PendingPods()
return len(pendingPods) == 3, nil
}); err != nil {
t.Fatal(err)
}
@ -263,7 +264,8 @@ func TestCustomResourceEnqueue(t *testing.T) {
// Wait for the testing Pod to be present in the scheduling queue.
if err := wait.Poll(time.Millisecond*200, wait.ForeverTestTimeout, func() (bool, error) {
return len(testCtx.Scheduler.SchedulingQueue.PendingPods()) == 1, nil
pendingPods, _ := testCtx.Scheduler.SchedulingQueue.PendingPods()
return len(pendingPods) == 1, nil
}); err != nil {
t.Fatal(err)
}