Scale MySQL on Kubernetes and OpenShift¶

One of the great advantages brought by Kubernetes and the OpenShift platform is the ease of an application scaling. Scaling an application results in adding resources or Pods and scheduling them to available Kubernetes nodes.

Scaling can be vertical and horizontal. Vertical scaling adds more compute or storage resources to MySQL nodes; horizontal scaling is about adding more nodes to the cluster.

Vertical scaling¶

Scale compute¶

There are multiple components that Operator deploys and manages: Percona XtraDB Cluster (PXC), HAProxy or ProxySQL, etc. To add or reduce CPU or Memory you need to edit corresponding sections in the Custom Resource. We follow the structure for requests and limits that Kubernetes provides .

To add more resources to your MySQL nodes in PXC edit the following section in the Custom Resource:

spec:
...
  pxc:
    ...
    resources:
      requests: 
        memory: 4G
        cpu: 2
      limits:
        memory: 4G
        cpu: 2

Use our reference documentation for the Custom Resource options for more details about other components.

Scale storage¶

Kubernetes manages storage with a PersistentVolume (PV), a segment of storage supplied by the administrator, and a PersistentVolumeClaim (PVC), a request for storage from a user. In Kubernetes v1.11 the feature was added to allow a user to increase the size of an existing PVC object (considered stable since Kubernetes v1.24). The user cannot shrink the size of an existing PVC object.

Starting from the version 1.14.0, the Operator allows to scale Percona XtraDB Cluster storage automatically by changing the appropriate Custom Resource option, if the volume type supports PVCs expansion.

Automated scaling with Volume Expansion capability¶

Certain volume types support PVCs expansion (exact details about PVCs and the supported volume types can be found in Kubernetes documentation ).

You can run the following command to check if your storage supports the expansion capability:

$ kubectl describe sc <storage class name> | grep allowVolumeExpansion

Expected output

allowVolumeExpansion: true

You can enable automated scaling with the enableVolumeExpansion Custom Resource option (turned off by default). When enabled, the Operator will automatically expand such storage for you when you change the pxc.volumeSpec.persistentVolumeClaim.resources.requests.storage option in the Custom Resource:

spec:
  ...
  enableVolumeExpansion: true
    ...
  pxc:
    ...
    volumeSpec:
      ...
      persistentVolumeClaim:
        resources:
          requests:
            storage: <NEW STORAGE SIZE>

Warning

If the new storage size can’t be reached because there is a resource quota in place and the PVC storage limits are reached, this will be detected, there will be no scaling attempts, and the Operator will revert the value in the Custom Resource option back. If resize isn’t successful (for example, no quota is set, but the new storage size turns out to be just too large), the Operator will detect Kubernetes failure on scaling, and revert the Custom Resource option. Still, Kubernetes will continue attempts to fulfill the scaling request until the problem is fixed manually by the Kubernetes administrator.

For example, you can do it by editing and applying the deploy/cr.yaml file:

spec:
...
  pxc:
    ...
    volumeSpec:
      persistentVolumeClaim:
        resources:
          requests:
            storage: <NEW STORAGE SIZE>

Apply changes as usual:

$ kubectl apply -f cr.yaml

Manual scaling without Volume Expansion capability¶

Manual scaling is the way to go if you version of the Operator is older than 1.14.0, your volumes have type which does not support Volume Expansion, or you just do not rely on automated scaling.

You will need to delete Pods one by one and their persistent volumes to resync the data to the new volumes. This can also be used to shrink the storage.

Update the Custom Resource with the new storage size by editing and applying the deploy/cr.yaml file:

spec:
...
  pxc:
    ...
    volumeSpec:
      persistentVolumeClaim:
        resources:
          requests:
            storage: <NEW STORAGE SIZE>

Apply the Custom Resource update in a usual way:

$ kubectl apply -f deploy/cr.yaml

Delete the StatefulSet with the orphan option
```
$ kubectl delete sts <statefulset-name> --cascade=orphan
```
The Pods will not go down and the Operator is going to recreate the StatefulSet:
```
$ kubectl get sts <statefulset-name>
```
Expected output
```
cluster1-pxc       3/3     39s
```

Scale up the cluster (Optional)

Changing the storage size would require us to terminate the Pods, which decreases the computational power of the cluster and might cause performance issues. To improve performance during the operation we are going to change the size of the cluster from 3 to 5 nodes:

...
spec:
...
  pxc:
    ...
    size: 5

Apply the change:

$ kubectl apply -f deploy/cr.yaml

New Pods will already have new storage:

$ kubectl get pvc

Expected output

NAME                     STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
datadir-cluster1-pxc-0   Bound    pvc-90f0633b-0938-4b66-a695-556bb8a9e943   10Gi       RWO            standard       110m
datadir-cluster1-pxc-1   Bound    pvc-7409ea83-15b6-448f-a6a0-12a139e2f5cc   10Gi       RWO            standard       109m
datadir-cluster1-pxc-2   Bound    pvc-90f0b2f8-9bba-4262-904c-1740fdd5511b   10Gi       RWO            standard       108m
datadir-cluster1-pxc-3   Bound    pvc-439bee13-3b57-4582-b342-98281aca50ba   19Gi       RWO            standard       49m
datadir-cluster1-pxc-4   Bound    pvc-2d4f3a60-4ec4-48a0-96cd-5243e2f05234   19Gi       RWO            standard       47m

Delete PVCs and Pods with old storage size one by one. Wait for data to sync before you proceeding to the next node.
```
$ kubectl delete pvc <PVC NAME>
$ kubectl delete pod <POD NAME>
```
The new PVC is going to be created along with the Pod.

Horizontal scaling¶

Size of the cluster is controlled by a size key in the Custom Resource options configuration. That’s why scaling the cluster needs nothing more but changing this option and applying the updated configuration file. This may be done in a specifically saved config:

spec:
...
  pxc:
    ...
    size: 5

Apply the change:

$ kubectl apply -f deploy/cr.yaml

Alternatively, you cana do it on the fly, using the following command:

$ kubectl scale --replicas=5 pxc/<CLUSTER NAME>

In this example we have changed the size of the Percona XtraDB Cluster to 5 instances.

Automated scaling¶

To automate horizontal scaling it is possible to use Horizontal Pod Autoscaler (HPA) . It will scale the Custom Resource itself, letting Operator to deal with everything else.

It is also possible to use Kuvernetes Event-driven Autoscaling (KEDA) , where you can apply more sophisticated logic for decision making on scaling.

For now it is not possible to use Vertical Pod Autoscaler (VPA) with the Operator due to the limitations it introduces for objects with owner references.

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