ebs br: make sure backup can be resumed even after being paused for a…

… long time (#5464)

pkg/backup/backupschedule/backup_schedule_manager.go

@@ -192,20 +192,18 @@ func getLastScheduledTime(bs *v1alpha1.BackupSchedule, nowFn nowFn) (*time.Time,
  // If there is a bug somewhere, or incorrect clock
  // on controller's server or apiservers (for setting creationTimestamp)
  // then there could be so many missed start times (it could be off
- // by decades or more), that it would eat up all the CPU and memory
- // of this controller. In that case, we want to not try to list
- // all the missed start times.
- //
- // I've somewhat arbitrarily picked 100, as more than 80,
- // but less than "lots".
- if len(scheduledTimes) > 100 {
+ // by decades or more). So, we need to set LastBackupTime to now() in order to let
+ // next reconcile succeed.
+ if len(scheduledTimes) > 1000 {
  // We can't get the last backup schedule time
  if bs.Status.LastBackupTime == nil && bs.Status.AllBackupCleanTime != nil {
  // Recovery backup schedule from pause status, should refresh AllBackupCleanTime to avoid unschedulable problem
  bs.Status.AllBackupCleanTime = &metav1.Time{Time: nowFn()}
  return nil, controller.RequeueErrorf("recovery backup schedule %s/%s from pause status, refresh AllBackupCleanTime.", ns, bsName)
  }
- klog.Error("Too many missed start backup schedule time (> 100). Check the clock.")
+ klog.Warning("Too many missed start backup schedule time (> 1000). Fail current one.")
+ offset := sched.Next(t).Sub(t)
+ bs.Status.LastBackupTime = &metav1.Time{Time: time.Now().Add(-offset)}
  return nil, nil
  }
  }

pkg/backup/backupschedule/backup_schedule_manager_test.go

@@ -176,10 +176,14 @@ func TestGetLastScheduledTime(t *testing.T) {
  }
 
  // test too many miss
- bs.Status.LastBackupTime.Time = now.AddDate(-1000, 0, 0)
+ bs.Status.LastBackupTime.Time = now.AddDate(-10, 0, 0)
  getTime, err = getLastScheduledTime(bs, time.Now)
  g.Expect(err).Should(BeNil())
  g.Expect(getTime).Should(BeNil())
+ // next reconcile should succeed
+ getTime, err = getLastScheduledTime(bs, time.Now)
+ g.Expect(err).Should(BeNil())
+ g.Expect(getTime).ShouldNot(BeNil())
 }
 
 func TestBuildBackup(t *testing.T) {

pkg/fedvolumebackup/backupschedule/backup_schedule_manager.go

@@ -117,20 +117,19 @@ func getLastScheduledTime(vbs *v1alpha1.VolumeBackupSchedule, nowFn nowFn) (*tim
  // If there is a bug somewhere, or incorrect clock
  // on controller's server or apiservers (for setting creationTimestamp)
  // then there could be so many missed start times (it could be off
- // by decades or more), that it would eat up all the CPU and memory
- // of this controller. In that case, we want to not try to list
- // all the missed start times.
- //
- // I've somewhat arbitrarily picked 100, as more than 80,
- // but less than "lots".
- if len(scheduledTimes) > 100 {
+ // by decades or more). So, we need to set LastBackupTime to now() in order to let
+ // next reconcile succeed.
+ if len(scheduledTimes) > 1000 {
  // We can't get the last backup schedule time
  if vbs.Status.LastBackupTime == nil && vbs.Status.AllBackupCleanTime != nil {
  // Recovery backup schedule from pause status, should refresh AllBackupCleanTime to avoid unschedulable problem
  vbs.Status.AllBackupCleanTime = &metav1.Time{Time: nowFn()}
  return nil, controller.RequeueErrorf("recovery backup schedule %s/%s from pause status, refresh AllBackupCleanTime.", ns, bsName)
  }
- klog.Error("Too many missed start backup schedule time (> 100). Check the clock.")
+
+ klog.Warning("Too many missed start backup schedule time (> 1000). Fail current one.")
+ offset := sched.Next(t).Sub(t)
+ vbs.Status.LastBackupTime = &metav1.Time{Time: time.Now().Add(-offset)}
  return nil, nil
  }
  }

pkg/fedvolumebackup/backupschedule/backup_schedule_manager_test.go

@@ -170,10 +170,14 @@ func TestGetLastScheduledTime(t *testing.T) {
  }
 
  // test too many miss
- bs.Status.LastBackupTime.Time = now.AddDate(-1000, 0, 0)
+ bs.Status.LastBackupTime.Time = now.AddDate(-10, 0, 0)
  getTime, err = getLastScheduledTime(bs, time.Now)
  g.Expect(err).Should(BeNil())
  g.Expect(getTime).Should(BeNil())
+ getTime, err = getLastScheduledTime(bs, time.Now)
+ // next reconcile should succeed
+ g.Expect(err).Should(BeNil())
+ g.Expect(getTime).ShouldNot(BeNil())
 }
 
 func TestBuildBackup(t *testing.T) {