Custom resource statuses

Status fields show the current state of a Custom Resource (CR). The Operator sets these fields in the .status section of a Custom Resource. You do not edit the status.

Use .status values to confirm progress, detect failures, and decide when it is safe to run the next action. For example, start a restore after a backup succeeds, or perform switchover after the ClusterSet is Ready.

How to view Custom Resource statuses

To check the status of your Percona Custom Resources, use the kubectl get <resource-type> or kubectl describe <resource-type> commands. See how to use them to get the quick overview, in-depth details, and targeted queries.

Get a quick overview

List your resources and check their high-level status:

kubectl get ps -n <namespace>
kubectl get ps-clusterset -n <namespace>
kubectl get ps-backup -n <namespace>
kubectl get ps-restore -n <namespace>

Sample output for PerconaServerMySQL

NAME          REPLICATION          ENDPOINT                              STATE   MYSQL   ORCHESTRATOR   HAPROXY   ROUTER   AGE
ps-cluster1   group-replication    ps-cluster1-haproxy.default           ready   3                      3                 27m

View full details

See all status details, conditions, and events:

kubectl get ps <cluster-name> -n <namespace> -o yaml
kubectl describe ps <cluster-name> -n <namespace>

kubectl get ps-clusterset <clusterset-name> -n <namespace> -o yaml
kubectl describe ps-clusterset <clusterset-name> -n <namespace>

kubectl get ps-backup <backup-name> -n <namespace> -o yaml
kubectl get ps-restore <restore-name> -n <namespace> -o yaml

Check for the .status field in the output to find the current state, readiness, messages, and conditions.

Query a status field directly

You can extract specific status fields using jsonpath.

Example 1. Get the InnoDB cluster name (for ClusterSet configuration):

kubectl get ps <cluster-name> -n <namespace> \
  -o jsonpath='{.status.innodbClusterName}{"\n"}'

Sample output

pscluster1

Example 2. Check whether the ClusterSet is ready:

kubectl get ps-clusterset <clusterset-name> -n <namespace> \
  -o jsonpath='{range .status.conditions[?(@.type=="Ready")]}{.status}{": "}{.reason}{"\n"}{end}'

Sample output

True: ClusterSetHealthy

Example 3. Get the backup destination path:

kubectl get ps-backup <backup-name> -n <namespace> \
  -o jsonpath='{.status.destination}{"\n"}'

Sample output

s3://<my-bucket>/ps-cluster1-2026-06-29-08:06:10-full

PerconaServerMySQL status

The main cluster state is recorded in status.state. Component-level states are recorded in the status.mysql, status.haproxy, status.router, status.orchestrator, and status.binlogserver sections. A component subsection is populated only when that component is enabled.

Common fields:

• status.state — overall cluster state
• status.host — connection endpoint (proxy Service, LoadBalancer address, or MySQL Service)
• status.innodbClusterName — InnoDB cluster name derived from the Custom Resource name (non-alphanumeric characters stripped). Use this value in the ClusterSet Custom Resource
• status.mysql.ready / status.mysql.size — number of ready MySQL Pods and the desired size
• status.mysql.state — MySQL component state
• status.haproxy / status.router / status.orchestrator / status.binlogserver — the same ready, size, state, and version fields for each component
• status.storageAutoscaling — per-PVC autoscaling progress when automatic storage scaling is enabled
• status.conditions — detailed condition list with reason and message. See Conditions

The overall status.state stays initializing until:

• every enabled component is ready,
• Group Replication members are ONLINE (for group-replication clusters) or asynchronous replication has no Orchestrator problems (for async clusters),
• point-in-time recovery’s binlog server is ready when enabled,
• any LoadBalancer Services have an assigned address.

Cluster state values

status.state values are:

Value	Meaning
""	The Operator has not set a state yet.
initializing	The Operator is creating or reconciling the cluster.
stopping	The Operator is stopping or scaling down cluster components (spec.pause: true and some Pods are still ready).
paused	The cluster is paused and no component Pods are ready.
ready	The cluster is up and healthy.
error	The Operator detected an error; check conditions and events.

The same state values apply to each component subsection (status.mysql.state, status.haproxy.state, and so on).

Conditions

Conditions show more detail about cluster state changes. You can see them in status.conditions[].

Common condition fields:

• type — condition type
• status — condition status
• reason — short reason string
• message — human-readable details

status.conditions[].type values:

Value	Meaning
Initializing	The cluster or a component is starting up.
Ready	The cluster or a component is ready.
Error	The Operator detected an error (for example, a reconcile failure or a full cluster crash).
InnoDBClusterBootstrapped	The InnoDB Cluster metadata exists and Group Replication is formed.
AwaitingExternalBootstrap	The cluster is configured with spec.mysql.bootstrap.mode: manual and waits for an external actor (typically the ClusterSet controller) to bootstrap Group Replication.
ClusterSetReplicationRunning	The cluster is a REPLICA member of an InnoDB ClusterSet and async replication from the primary cluster is active.

status.conditions[].status values:

Value	Meaning
True	The condition is currently true.
False	The condition is currently false.
Unknown	The Operator has not determined the condition yet.

The Operator sets reason and message as free-form strings. Common reasons include:

• Initializing, Ready (matching the cluster state)
• ErrorReconcile — cluster reconciliation failed; see the condition message
• FullClusterCrashDetected — all Group Replication members have crash recovery files
• ManualBootstrapRequested — the cluster is waiting for external bootstrap
• ClusterSetReplicationRunning — ClusterSet replica replication is active
• InnoDBClusterBootstrapped — InnoDB Cluster metadata exists

Storage autoscaling status

When you enable storage autoscaling, the Operator records per-PVC progress in status.storageAutoscaling.<pvc-name>:

Field	Meaning
currentSize	Observed PVC size
lastResizeTime	Timestamp of the last expansion
resizeCount	Number of expansions
lastError	Last autoscaling error, if any

Events

The cluster controller emits Kubernetes events you can view with kubectl describe ps:

Event reason	Meaning
ClusterStateChanged	status.state changed (for example, initializing → ready).
ReconcileError	Cluster reconciliation failed.
FullClusterCrashDetected	A full Group Replication cluster crash was detected.
AsyncReplicationNotReady	Orchestrator reported replication problems on one or more instances.
StorageAutoscalingTriggered	Storage autoscaling started a PVC resize.

PerconaServerMySQLClusterSet status

ClusterSet progress and topology are reflected in status.primaryCluster, status.clusters, and status.conditions. For configuration options, see ClusterSet Resource options. For setup, see Set up cross-site replication.

Common fields:

• status.primaryCluster — the cluster currently serving as the ClusterSet primary (observed; may lag spec.primaryCluster briefly during switchover)
• status.primaryClusterEndpoint — host:port of the global primary instance
• status.clusters — map of member clusters keyed by innodbClusterName, mirroring MySQL Shell .status() output
• status.conditions — ClusterSet-level conditions
• status.lastObservedAt — timestamp of the last successful status refresh
• status.lastObservedGeneration — generation last reconciled

Per-member cluster status

Each entry in status.clusters[<innodbClusterName>] contains:

Field	Meaning
clusterRole	PRIMARY or REPLICA as observed by MySQL Shell
globalStatus	Member health in the ClusterSet
primary	Host:port of the cluster’s local Group Replication primary

globalStatus values (from MySQL Shell):

Value	Meaning
OK	Cluster is healthy and replicating as expected.
OK_NOT_REPLICATING	Cluster is reachable but not currently replicating (transitional or misconfigured).
NOT_OK	Cluster has a problem; check MySQL Shell status and events.
INVALIDATED	Cluster was fenced off after forced failover; it may have divergent GTIDs. Rejoin it with rejoinCluster() if GTIDs are compatible. Otherwise remove and recreate the cluster to rejoin the ClusterSet.
UNKNOWN	Status could not be determined (for example, the primary is unreachable).

Conditions

status.conditions[].type values:

Value	Meaning
Ready	ClusterSet is formed and replicas are replicating from the specified primary.
ClusterSetBootstrapped	The primary cluster is configured as a ClusterSet.
MySQLShellRunnerReady	The mysqlshell-runner Pod is running and ready.
SwitchoverInProgress	spec.primaryCluster differs from the observed primary and a switchover Job is running, pending, or failed.
ErrorReconcile	An error occurred during reconciliation.
ReplicaManagementFailure	A replica add or remove Job failed; see the Job logs.
ClusterSetDissolving	The Custom Resource is being deleted and the dissolve finalizer is running.

Common Ready condition reasons:

Reason	Meaning
ClusterSetHealthy	MySQL Shell reports overall ClusterSet status as healthy.
ClusterSetNotHealthy	One or more members are not healthy; see status.clusters and the condition message.
PrimaryUnreachable	Primary cluster is not reachable. Ready is Unknown.
AccessDenied	Incorrect password configured on the replica site. Ready is Unknown.

Common MySQLShellRunnerReady reasons:

Reason	Meaning
DeploymentReady	The mysqlshell-runner Deployment is current and ready.
DeploymentNotReady	The Deployment is still rolling out.
DeploymentFailed	The Deployment failed.
DeploymentNotObserved	The Operator has not observed Deployment status yet.

Common SwitchoverInProgress reasons:

Reason	Meaning
SwitchoverInProgress	A switchover Job is running. Condition status is True.
SwitchoverFailed	The switchover Job failed. Condition status is False.

ErrorReconcile uses the reason ErrorReconcile for general failures, or AccessDenied / PrimaryUnreachable when the ClusterSet manager cannot reach or authenticate to the primary.

Events

The ClusterSet controller emits Kubernetes events you can view with kubectl describe ps-clusterset:

Event reason	Meaning
ClusterSetBootstrapped	ClusterSet was created on the primary cluster.
ClusterSetPrimarySwitched	Planned switchover completed.
ClusterSetPrimaryForcedSwitched	Forced failover completed.
ClusterSetMemberAdded	A cluster was added to the ClusterSet.
ClusterSetMemberRemoved	A cluster was removed from the ClusterSet.
ClusterSetHealthDegraded	Overall ClusterSet health dropped from healthy to unhealthy.

PerconaServerMySQLBackup status

Backup progress and results are in status.state. You also get destination and timing details that help you validate backups and incremental chains.

Common fields:

• status.state — backup job state
• status.type — backup type (full or incremental)
• status.destination — backup path or URL (s3://, gs://, or Azure container path)
• status.completed — completion timestamp
• status.stateDescription — error details when the backup fails
• status.backupSource — source Pod used for the backup
• status.storage — copy of the storage configuration used for this backup
• status.image — Percona XtraBackup image used for the Job
• status.compressed — whether the backup was compressed
• status.conditions — additional conditions

Backup state values

status.state values are:

Value	Meaning
""	Backup is created but not processed yet.
Starting	Backup is starting.
Running	Backup is in progress.
Succeeded	Backup completed successfully.
Error	Backup failed to start (for example, the cluster is not ready, backups are disabled, or the storage name is missing). Check status.stateDescription. This state is terminal; fix the cause and create a new backup.
Failed	Backup started but failed during execution. Check status.stateDescription and the backup Job logs.

For troubleshooting steps, see Troubleshoot backups and restores.

Backup conditions

PerconaServerMySQLBackup may include the following condition:

Value	Meaning
BackupLeaseAcquired	The backup Job holds the cluster backup lease and is authorized to run. When False with reason LeaseAlreadyHeld, another backup still holds the lease.

PerconaServerMySQLRestore status

Restore progress and results are in status.state. Use these fields to confirm when a restore starts, finishes, or fails.

Common fields:

• status.state — restore job state
• status.stateDescription — error details when the restore cannot proceed or fails
• status.completed — completion timestamp

Restore state values

status.state values are:

Value	Meaning
""	Restore is created but not processed yet.
Starting	Restore is starting.
Running	Restore is in progress, including waiting for the cluster to become ready after the restore Job completes.
Succeeded	Restore completed successfully.
Error	Restore cannot start yet. Typical causes: the cluster is not found, backupName and backupSource are empty, backupSource is incomplete, the storage name is missing from the cluster Custom Resource, or point-in-time recovery is requested but the binlog server is not enabled. Check status.stateDescription. The Operator retries on the next reconcile.
Failed	The restore Job, prepare Job, or point-in-time recovery Job failed. Check the Job logs.

If another restore holds the cluster lease, or a backup is still running, status.state stays empty ("") and status.stateDescription explains the wait. That is not the Error state.

Error vs Failed

• Error — Recoverable. The Operator sets this state when validation fails or a prerequisite is missing. It resets the state on the next reconcile, so you can fix the Restore object configuration and unblock the restore process without creating a new Restore object.
• Failed — Terminal. The restore, prepare, or point-in-time recovery Job failed. Inspect the Job logs, fix the cause, and create a new Restore Custom Resource.

When a restore is in Error, inspect the message:

kubectl get ps-restore <restore-name> -n <namespace> \
  -o jsonpath='{.status.state}{"\n"}{.status.stateDescription}{"\n"}'

---

Last update: September 1, 2026
Created: September 1, 2026