Set up Percona XtraDB Cluster cross-site replication

The cross-site replication involves configuring one Percona XtraDB Cluster as Source, and another Percona XtraDB Cluster as Replica to allow an asynchronous replication between them:

The Operator automates configuration of Source and Replica Percona XtraDB Clusters, but the feature itself is not bound to Kubernetes. Either Source or Replica can run outside of Kubernetes, be regular MySQL and be out of the Operators’ control.

This feature can be useful in several cases: for example, it can simplify migration from on-premises to the cloud with replication, and it can be really helpful in case of the disaster recovery too.

Note

Cross-site replication is based on Automatic Asynchronous Replication Connection Failover . Therefore it requires MySQL 8.0.22+ (Percona XtraDB Cluster 8.0.22+) to work.

Setting up MySQL for asynchronous replication without the Operator is out of the scope for this document, but it is described here and is also covered by this HowTo.

Configuring the cross-site replication for the cluster controlled by the Operator is explained in the following subsections.

Creating a Replica cluster

Cross-site replication can be configured on two sibling Percona XtraDB Clusters. That’s why you should first make a fully operational clone of your main database cluster. After that your original cluster will be configured as Source, and a new one (the clone) will be configured as Replica.

The easiest way to achieve this is to use backups. You make a full backup of your main database cluster, and restore it to a new Kubernetes-based environment, following this HowTo.

Configuring cross-site replication on Source instances

You can configure Source instances for cross-site replication with spec.pxc.replicationChannels subsection in the deploy/cr.yaml configuration file. It is an array of channels, and you should provide the following keys for the channel in your Source Percona XtraDB Cluster:

• pxc.replicationChannels.[].name key is the name of the channel,
• pxc.replicationChannels.[].isSource key should be set to true.

Here is an example:

spec:
  pxc:
    replicationChannels:
    - name: pxc1_to_pxc2
      isSource: true

You will also need to expose every Percona XtraDB Cluster Pod of the Source cluster to make it possible for the Replica cluster to connect. This is done through the pxc.expose section in the deploy/cr.yaml configuration file as follows.

spec:
  pxc:
    expose:
      enabled: true
      type: LoadBalancer

Note

This will create a LoadBalancer per each Percona XtraDB Cluster Pod. In most cases, for cross-region replication to work this Load Balancer should be internet-facing.

The cluster will be ready for asynchronous replication when you apply changes as usual:

$ kubectl apply -f deploy/cr.yaml

To list the endpoints assigned to PXC Pods list the Kubernetes Service objects by executing kubectl get services -l "app.kubernetes.io/instance=cluster1" command (don’t forget to substitute cluster1 with the real name of your cluster, if you don’t use the default name).

Configuring cross-site replication on Replica instances

You can configure Replica instances for cross-site replication with spec.pxc.replicationChannels subsection in the deploy/cr.yaml configuration file. It is an array of channels, and you should provide the following keys for the channel in your Replica Percona XtraDB Cluster:

• pxc.replicationChannels.[].name key is the name of the channel,
• pxc.replicationChannels.[].isSource key should be set to false,
• pxc.replicationChannels.[].sourcesList is the list of Source cluster names from which Replica should get the data,
• pxc.replicationChannels.[].sourcesList.[].host is the host name or IP address of the Source,
• pxc.replicationChannels.[].sourcesList.[].port is the port of the source (3306 port will be used if nothing specified),
• pxc.replicationChannels.[].sourcesList.[].weight is the weight of the source (in the event of a connection failure, a new source is selected from the list based on a weighted priority).

Here is the example:

spec:
  pxc:
    replicationChannels:
    - name: uspxc1_to_pxc2
      isSource: false
      sourcesList:
      - host: pxc1.source.percona.com
        port: 3306
        weight: 100
      - host: pxc2.source.percona.com
        weight: 100
      - host: pxc3.source.percona.com
        weight: 100
    - name: eu_to_pxc2
      isSource: false
      sourcesList:
      - host: pxc1.source.percona.com
        port: 3306
        weight: 100
      - host: pxc2.source.percona.com
        weight: 100
      - host: pxc3.source.percona.com
        weight: 100

The cluster will be ready for asynchronous replication when you apply changes as usual:

$ kubectl apply -f deploy/cr.yaml

Note

You can also configure SSL channel for replication . Following options allow you using replication over an encrypted channel. Set the replicationChannels.configuration.ssl key to true, optionally enable host name identity verification with the replicationChannels.configuration.sslSkipVerify key, and set replicationChannels.configuration.ca key to the path name of the Certificate Authority (CA) certificate file:

replicationChannels:
- isSource: false
  name: uspxc1_to_pxc2
  configuration:
    ssl: true
    sslSkipVerify: true
    ca: '/etc/mysql/ssl/ca.crt'
    ...

SSL certificates on both sides should be signed by the same certificate authority for encrypted replication channels to work.

System user for replication

Replication channel demands a special system user with same credentials on both Source and Replica.

The Operator creates a system-level Percona XtraDB Cluster user named replication for this purpose, with credentials stored in a Secret object along with other system users.

Note

If the Replica cluster is not a clone of the original one (for example, it’s outside of Kubernetes and is not under the Operator’s control) the appropriate user with necessary permissions should be created manually.

If you need you can change a password for this user as follows:

in Linuxin macOS

$ kubectl patch secret/cluster1-secrets -p '{"data":{"replication": "'$(echo -n new_password | base64 --wrap=0)'"}}'

$ kubectl patch secret/cluster1-secrets -p '{"data":{"replication": "'$(echo -n new_password | base64)'"}}'

If you have changed the replication user’s password on the Source cluster, and you use the Operator version 1.9.0, you can have a replication is not running error message in log, similar to the following one:

{"level":"info","ts":1629715578.2569592,"caller":"zapr/zapr.go 69","msg":"Replication for channel is not running. Please, check the replication status","channel":"pxc2_to_pxc1"}

Fixing this involves the following steps.

1. Find the Replica Pod which was chosen by the Operator for replication, using the following command:

   $ kubectl get pods --selector percona.com/replicationPod=true
   

2. Get the shell access to this Pod and login to the MySQL monitor as a root user:

   $ kubectl exec -c pxc --stdin --tty <pod_name> -- /bin/bash
   bash-4.4$ mysql -uroot -proot_password
   

3. Execute the following three SQL commands to propagate the replication user password from the Source cluster to Replica:

   STOP REPLICA IO_THREAD FOR CHANNEL '$REPLICATION_CHANNEL_NAME';
   CHANGE MASTER TO MASTER_PASSWORD='$NEW_REPLICATION_PASSWORD' FOR CHANNEL '$REPLICATION_CHANNEL_NAME';
   START REPLICA IO_THREAD FOR CHANNEL '$REPLICATION_CHANNEL_NAME';
   

Get expert help

If you need assistance, visit the community forum for comprehensive and free database knowledge, or contact our Percona Database Experts for professional support and services. Join K8S Squad to benefit from early access to features and “ask me anything” sessions with the Experts.

Community Forum Get a Percona Expert Join K8S Squad

---

Last update: 2024-10-29