public_git
/
postgres-operator
mirror of https://github.com/zalando/postgres-operator.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

wait only after restart not every sync

This commit is contained in:
Felix Kunde 2021-10-22 10:07:42 +02:00
parent 0b32b42f0b
commit 4174cd6639
1 changed files with 7 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
pkg/cluster/sync.go
View File

@ -273,10 +273,9 @@ func (c *Cluster) syncPodDisruptionBudget(isUpdate bool) error {
func (c *Cluster) syncStatefulSet() error { func (c *Cluster) syncStatefulSet() error {
var ( var (
restartTTL uint32 restartWait uint32
restartMasterFirst bool restartMasterFirst bool
) )
podsToRecreate := make([]v1.Pod, 0) podsToRecreate := make([]v1.Pod, 0)
switchoverCandidates := make([]spec.NamespacedName, 0) switchoverCandidates := make([]spec.NamespacedName, 0)
@ -410,6 +409,7 @@ func (c *Cluster) syncStatefulSet() error {
c.logger.Warningf("could not get Postgres config from pod %s: %v", podName, err) c.logger.Warningf("could not get Postgres config from pod %s: %v", podName, err)
continue continue
} }
restartWait = patroniConfig.LoopWait
// empty config probably means cluster is not fully initialized yet, e.g. restoring from backup // empty config probably means cluster is not fully initialized yet, e.g. restoring from backup
// do not attempt a restart // do not attempt a restart
@ -420,8 +420,7 @@ func (c *Cluster) syncStatefulSet() error {
continue continue
} }
// it could take up to LoopWait to apply the config // it could take up to LoopWait to apply the config
time.Sleep(time.Duration(patroniConfig.LoopWait)*time.Second + time.Second*2) time.Sleep(time.Duration(restartWait)*time.Second + time.Second*2)
restartTTL = patroniConfig.TTL
break break
} }
} }
@ -438,19 +437,13 @@ func (c *Cluster) syncStatefulSet() error {
remainingPods = append(remainingPods, &pods[i]) remainingPods = append(remainingPods, &pods[i])
continue continue
} }
c.restartInstance(&pod) c.restartInstance(&pod, restartWait)
if len(pods) > 1 {
time.Sleep(time.Duration(restartTTL) * time.Second)
}
} }
// in most cases only the master should be left to restart // in most cases only the master should be left to restart
if len(remainingPods) > 0 { if len(remainingPods) > 0 {
for _, remainingPod := range remainingPods { for _, remainingPod := range remainingPods {
c.restartInstance(remainingPod) c.restartInstance(remainingPod, restartWait)
if len(remainingPods) > 1 {
time.Sleep(time.Duration(restartTTL) * time.Second)
}
} }
} }
@ -467,7 +460,7 @@ func (c *Cluster) syncStatefulSet() error {
return nil return nil
} }
func (c *Cluster) restartInstance(pod *v1.Pod) { func (c *Cluster) restartInstance(pod *v1.Pod, restartWait uint32) {
podName := util.NameFromMeta(pod.ObjectMeta) podName := util.NameFromMeta(pod.ObjectMeta)
role := PostgresRole(pod.Labels[c.OpConfig.PodRoleLabel]) role := PostgresRole(pod.Labels[c.OpConfig.PodRoleLabel])
@ -485,6 +478,7 @@ func (c *Cluster) restartInstance(pod *v1.Pod) {
c.logger.Warningf("could not restart Postgres server within %s pod %s: %v", role, podName, err) c.logger.Warningf("could not restart Postgres server within %s pod %s: %v", role, podName, err)
return return
} }
time.Sleep(time.Duration(restartWait) * time.Second)
c.eventRecorder.Event(c.GetReference(), v1.EventTypeNormal, "Update", fmt.Sprintf("Postgres server restart done for %s pod %s", role, pod.Name)) c.eventRecorder.Event(c.GetReference(), v1.EventTypeNormal, "Update", fmt.Sprintf("Postgres server restart done for %s pod %s", role, pod.Name))
} }
} }