Objects the Operator creates¶

This page lists what the Operator creates in your Kubernetes cluster for a Percona XtraDB Cluster, and which containers run in each Pod. Use it to size quotas, plan RBAC and storage, and to find an object when you troubleshoot.

Kubernetes objects¶

For a cluster named cluster1, the Operator creates and manages the following Kubernetes objects. Some exist only when the related feature is turned on.

Kind	Name	Purpose
StatefulSet	`cluster1-pxc`	Database nodes
StatefulSet	`cluster1-haproxy` or `cluster1-proxysql`	Proxy, depending on which one is enabled
Service	`cluster1-haproxy`, `cluster1-proxysql`, `cluster1-pxc`, and more	Client entry points, headless Services for node discovery, and optional per-Pod Services. See Services and ports
Deployment and Service	`cluster1-pitr`	Binlog collector, only when point-in-time recovery is turned on
Secret	`cluster1-secrets`	Passwords of system users. You can edit it to change passwords
Secret	`internal-cluster1`	Passwords the Operator has actually applied to the database. Do not edit it
Secret	`cluster1-ssl`, `cluster1-ssl-internal`	TLS certificates for client and internal traffic
ConfigMap	`cluster1-pxc`, `cluster1-haproxy`, `cluster1-proxysql`	Your custom configuration from the CR
ConfigMap	`auto-cluster1-pxc`	MySQL options the Operator calculates from the memory limit
PodDisruptionBudget	one per component	Limits voluntary disruptions, such as node drains, to one Pod at a time by default
PersistentVolumeClaim	`datadir-cluster1-pxc-N`, `proxydata-cluster1-proxysql-N`	Data volumes, created by the StatefulSets
Lease	`pxc-cluster1-backup-lock`	Makes backups run one at a time. Created only when `backup.allowParallel` is `false`
Job	`xb-<backup>`	Makes a backup and uploads it to storage. Created for each backup
Job	`restore-job-<restore>-cluster1`	Writes the backup into the data volume of `cluster1-pxc-0`. Created for each restore
Job	`prepare-job-<restore>-cluster1`	Prepares the restored data so that the node can bootstrap a new cluster. Created for each restore
Job	`pitr-job-<restore>-cluster1`	Replays binlogs up to the target. Created only for a point-in-time restore

If cert-manager is installed, the Operator also creates cert-manager Issuers and Certificates that produce the TLS Secrets.

Note

Scheduled backups are not Kubernetes CronJobs. The Operator keeps the schedule itself and creates a backup object when a run is due. See Scheduled backups.

Objects that survive cluster deletion

PVCs and Secrets are not owned by the Custom Resource, so deleting the CR does not delete them. This lets you recreate the cluster with its data and credentials. To delete them together with the cluster, set the percona.com/delete-pxc-pvc and percona.com/delete-ssl finalizers. ProxySQL volumes have their own finalizer, percona.com/delete-proxysql-pvc. See Delete the database cluster.

Containers in each Pod¶

Percona XtraDB Cluster Pod (cluster1-pxc-N):

pxc-init (init container) copies the Operator’s entrypoint, probe and helper scripts into the data volume.
pxc runs mysqld. It listens on 3306 (MySQL), 33062 (admin), 33060 (MySQL X protocol), and on 4567, 4568 and 4444 for Galera replication, IST and SST.
Optional sidecars: logs and logrotate for log collection, pmm-client for monitoring, xtrabackup when the XtrabackupSidecar feature gate is on, and your own sidecar containers.

HAProxy Pod (cluster1-haproxy-N):

haproxy routes client traffic.
pxc-monit watches the cluster1-pxc Service and regenerates the HAProxy configuration when nodes join or leave.
Optional: pmm-client.

ProxySQL Pod (cluster1-proxysql-N):

proxysql routes client traffic.
pxc-monit watches the cluster1-pxc Service and adds or removes database nodes in ProxySQL.
proxysql-monit watches the cluster1-proxysql-unready Service so that ProxySQL instances form a ProxySQL cluster. It is not created when the ProxySQL scheduler is enabled.
Optional: pmm-client.

Objects the Operator creates¶

Kubernetes objects¶

Containers in each Pod¶

See also¶