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

fix migration of single-node clusters (#2134)

This commit is contained in:
Felix Kunde 2022-12-09 12:42:10 +01:00 committed by GitHub
parent 4d585250db
commit 0bef3b325f
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 10 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
pkg/cluster/pod.go
View File

@ -214,19 +214,16 @@ func (c *Cluster) movePodFromEndOfLifeNode(pod *v1.Pod) (*v1.Pod, error) {
// MigrateMasterPod migrates master pod via failover to a replica // MigrateMasterPod migrates master pod via failover to a replica
func (c *Cluster) MigrateMasterPod(podName spec.NamespacedName) error { func (c *Cluster) MigrateMasterPod(podName spec.NamespacedName) error {
var ( var (
masterCandidateName spec.NamespacedName
err error err error
eol bool eol bool
) )
oldMaster, err := c.KubeClient.Pods(podName.Namespace).Get(context.TODO(), podName.Name, metav1.GetOptions{}) oldMaster, err := c.KubeClient.Pods(podName.Namespace).Get(context.TODO(), podName.Name, metav1.GetOptions{})
if err != nil { if err != nil {
return fmt.Errorf("could not get pod: %v", err) return fmt.Errorf("could not get master pod: %v", err)
} }
c.logger.Infof("starting process to migrate master pod %q", podName) c.logger.Infof("starting process to migrate master pod %q", podName)
if eol, err = c.podIsEndOfLife(oldMaster); err != nil { if eol, err = c.podIsEndOfLife(oldMaster); err != nil {
return fmt.Errorf("could not get node %q: %v", oldMaster.Spec.NodeName, err) return fmt.Errorf("could not get node %q: %v", oldMaster.Spec.NodeName, err)
} }
@ -250,20 +247,20 @@ func (c *Cluster) MigrateMasterPod(podName spec.NamespacedName) error {
} }
c.Statefulset = sset c.Statefulset = sset
} }
// We may not have a cached statefulset if the initial cluster sync has aborted, revert to the spec in that case. // we may not have a cached statefulset if the initial cluster sync has aborted, revert to the spec in that case
masterCandidateName := podName
masterCandidatePod := oldMaster
if *c.Statefulset.Spec.Replicas > 1 { if *c.Statefulset.Spec.Replicas > 1 {
if masterCandidateName, err = c.getSwitchoverCandidate(oldMaster); err != nil { if masterCandidateName, err = c.getSwitchoverCandidate(oldMaster); err != nil {
return fmt.Errorf("could not find suitable replica pod as candidate for failover: %v", err) return fmt.Errorf("could not find suitable replica pod as candidate for failover: %v", err)
} }
} else { masterCandidatePod, err = c.KubeClient.Pods(masterCandidateName.Namespace).Get(context.TODO(), masterCandidateName.Name, metav1.GetOptions{})
c.logger.Warningf("migrating single pod cluster %q, this will cause downtime of the Postgres cluster until pod is back", c.clusterName())
}
masterCandidatePod, err := c.KubeClient.Pods(masterCandidateName.Namespace).Get(context.TODO(), masterCandidateName.Name, metav1.GetOptions{})
if err != nil { if err != nil {
return fmt.Errorf("could not get master candidate pod: %v", err) return fmt.Errorf("could not get master candidate pod: %v", err)
} }
} else {
c.logger.Warningf("migrating single pod cluster %q, this will cause downtime of the Postgres cluster until pod is back", c.clusterName())
}
// there are two cases for each postgres cluster that has its master pod on the node to migrate from: // there are two cases for each postgres cluster that has its master pod on the node to migrate from:
// - the cluster has some replicas - migrate one of those if necessary and failover to it // - the cluster has some replicas - migrate one of those if necessary and failover to it