fix(resources duration): handle failed pod started time set to unix e…

…poch

We have observed that containerd somtimes fails a pod on GKE and sets the `startedAt` value to
"1970-01-01T00:00:00Z" which causes argo workflows to calculate invalid resources duration.

fix #13709

util/resource/summary.go

@@ -8,13 +8,17 @@ import (
  wfv1 "github.com/argoproj/argo-workflows/v3/pkg/apis/workflow/v1alpha1"
 )
 
+func unixEpoch() time.Time {
+ return time.Unix(0, 0).UTC()
+}
+
 type Summary struct {
  ResourceList corev1.ResourceList
  ContainerState corev1.ContainerState
 }
 
 func (s Summary) age() time.Duration {
- if s.ContainerState.Terminated != nil {
+ if s.ContainerState.Terminated != nil && s.ContainerState.Terminated.StartedAt.Time != unixEpoch() {
  return s.ContainerState.Terminated.FinishedAt.Time.Sub(s.ContainerState.Terminated.StartedAt.Time)
  } else {
  return 0

util/resource/summary_test.go

@@ -0,0 +1,73 @@
+package resource
+
+import (
+ "testing"
+ "time"
+
+ wfv1 "github.com/argoproj/argo-workflows/v3/pkg/apis/workflow/v1alpha1"
+ "github.com/stretchr/testify/assert"
+ "k8s.io/apimachinery/pkg/api/resource"
+ metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+
+ corev1 "k8s.io/api/core/v1"
+)
+
+func TestSummaries_Duration(t *testing.T) {
+ startTime := time.Now().Add(-1 * time.Hour)
+ finishTime := time.Now()
+
+ resourceList := corev1.ResourceList{
+ corev1.ResourceCPU: resource.MustParse("1"),
+ corev1.ResourceMemory: resource.MustParse("1Gi"),
+ }
+
+ summaries := Summaries{
+ "container1": {
+ ResourceList: resourceList,
+ ContainerState: corev1.ContainerState{
+ Terminated: &corev1.ContainerStateTerminated{
+ StartedAt: metav1.NewTime(startTime),
+ FinishedAt: metav1.NewTime(finishTime),
+ },
+ },
+ },
+ }
+
+ expectedDuration := wfv1.ResourcesDuration{}
+ expectedDuration = expectedDuration.Add(wfv1.ResourcesDuration{
+ corev1.ResourceCPU: wfv1.NewResourceDuration(3600 * time.Second),
+ corev1.ResourceMemory: wfv1.NewResourceDuration(36864 * time.Second),
+ })
+
+ assert.Equal(t, expectedDuration, summaries.Duration())
+}
+
+func TestSummaries_Duration_StartedAtIsEpoch(t *testing.T) {
+ startedTime, err := time.Parse(time.RFC3339, "1970-01-01T00:00:00Z")
+ assert.NoError(t, err)
+ finishTime := time.Now()
+
+ resourceList := corev1.ResourceList{
+ corev1.ResourceCPU: resource.MustParse("1"),
+ corev1.ResourceMemory: resource.MustParse("1Gi"),
+ }
+
+ summaries := Summaries{
+ "container1": {
+ ResourceList: resourceList,
+ ContainerState: corev1.ContainerState{
+ Terminated: &corev1.ContainerStateTerminated{
+ StartedAt: metav1.NewTime(startedTime),
+ FinishedAt: metav1.NewTime(finishTime),
+ },
+ },
+ },
+ }
+
+ expectedDuration := wfv1.ResourcesDuration{}
+ expectedDuration = expectedDuration.Add(wfv1.ResourcesDuration{
+ corev1.ResourceCPU: wfv1.NewResourceDuration(0 * time.Second),
+ corev1.ResourceMemory: wfv1.NewResourceDuration(0 * time.Second),
+ })
+ assert.Equal(t, expectedDuration, summaries.Duration())
+}