Glossary¶
This page explains the Kubernetes and Galera terms used in the Operator documentation. The examples use a cluster named cluster1.
Arbitrator (garbd)¶
The Galera Arbitrator daemon, garbd. A member of the Galera group that doesn’t run a database server. In the default backup method, the backup Job joins the cluster as an arbitrator to receive a copy of the data through SST.
See Backup methods.
Cluster¶
The word cluster has two meanings in these docs. A Kubernetes cluster is the set of machines that Kubernetes manages. A Percona XtraDB Cluster is the database cluster that one Custom Resource describes: its database nodes and its proxies. One Kubernetes cluster can run several of them. Pages that say “the cluster” mean the database cluster, unless they say “Kubernetes cluster”.
Custom Resource (CR, cr.yaml)¶
The Kubernetes object that describes the cluster you want, for example PerconaXtraDBCluster. The Operator reads it and makes the cluster match. The CustomResourceDefinition (CRD) tells Kubernetes that this object type exists.
The deploy/cr.yaml file in the Operator repository is the example manifest of this object. See Change the cluster through the Custom Resource.
For all options, see the Custom Resource options reference.
Donor¶
The database node that sends data to a node that joins or rejoins the cluster, through IST or SST. Galera chooses the donor.
See In-cluster failover.
Finalizer¶
A key in metadata.finalizers that makes Kubernetes wait for a cleanup step before it deletes an object. The Operator uses finalizers such as percona.com/delete-pxc-pvc to clean up volumes and certificates.
See Clean up resources.
Galera¶
The replication library in Percona XtraDB Cluster. It replicates transactions to every node, keeps quorum and brings joining nodes up to date. The Operator doesn’t control it at runtime.
See Who owns what.
gcache¶
The Galera write-set cache on each node. If the missing transactions are still in the donor’s gcache, a joining node catches up with IST. If not, it needs SST.
See In-cluster failover.
Headless Service¶
A Kubernetes Service without a cluster IP. DNS returns the addresses of the Pods behind it. The database nodes and the proxies use headless Services such as cluster1-pxc to find each other.
See Services and ports.
Incremental State Transfer (IST)¶
A transfer of only the transactions a joining node missed, taken from the donor’s gcache. Compare SST.
See In-cluster failover.
Lease¶
A small Kubernetes object that works as a lock. The Operator uses a Lease so that only one backup runs at a time when backup.allowParallel is false.
See Objects the Operator creates.
Node¶
The word node has two meanings in these docs. A Kubernetes node is a machine, physical or virtual, where Pods run. A database node is one member of the Percona XtraDB Cluster: a cluster1-pxc-N Pod that runs mysqld. By default, the Operator requires each database Pod to run on a different Kubernetes node.
PersistentVolumeClaim (PVC)¶
A request for storage. Each database Pod has a PVC that holds its data. The StorageClass and the Container Storage Interface (CSI) driver provide the actual volume.
See Resize storage.
PodDisruptionBudget (PDB)¶
A Kubernetes object that limits how many Pods of a component can be unavailable during voluntary disruptions such as a node drain. The Operator creates one for each component.
Primary component¶
The group of nodes that has quorum and serves queries. A node outside it reports non-Primary and rejects queries.
Quorum¶
A majority of the nodes. A cluster of three nodes keeps quorum when one node is lost. It stops serving when it loses a second one.
Reconcile¶
One pass of the Operator loop: it compares the Custom Resource with what exists in Kubernetes and corrects the difference. The Operator repeats it about every five seconds.
See How the Operator reacts to changes.
seqno¶
The sequence number of the last transaction a node committed. After a full cluster crash, the Operator starts recovery from the node with the highest seqno.
See Crash recovery.
Smart Update¶
The Operator update strategy that restarts the database nodes one at a time and keeps quorum. It doesn’t start while a backup is running.
State Snapshot Transfer (SST)¶
A full copy of the data from a donor to a joining node, made with Percona XtraBackup. Compare IST.
See In-cluster failover.
StatefulSet¶
A Kubernetes workload that gives each Pod a stable name and its own volume, for example cluster1-pxc-0. The Operator runs the database nodes and the proxies as StatefulSets.
See Objects the Operator creates.
Created: November 3, 2026