Expanding Memphis Disk Storage

Kubernetes is commonly employed to deploy stateful applications like Memphis within your cluster. Each Pod in the StatefulSets retains access to its local persistent volumes even after rescheduling, ensuring the individual state is maintained, separate from neighboring Pods within the set.

However, these volumes have a notable limitation: Kubernetes does not offer a straightforward way to resize them through the StatefulSet object. The spec.resources.requests.storage property within the volumeClaimTemplates field of the StatefulSet is immutable and restricts you from making necessary capacity increases. This article will demonstrate how to work around this limitation.

Step 1: Verify Dynamic Resize Availability

Start by confirming if dynamic resizing is available in the storage class. If not, add the following line to enable it:

allowVolumeExpansion: true

Step 2: Increase PVC Size

To increase the size of the attached PVC (repeat for all PVC instances), execute the following command:

kubectl patch pvc  memphis-js-pvc-memphis-0 -p '{"spec":{"resources":{"requests":{"storage":"600Gi"}}}}' -n memphis

Step 3: Update StatefulSet Configuration

Export the Memphis StatefulSet configuration to a YAML file and adjust the size of the PVC accordingly:

k get sts memphis -n memphis -o yaml > sts.yaml

Edit the file to reflect the new PVC size within the spec section:

    spec:
      accessModes:
      - ReadWriteOnce
      resources:
        requests:
          storage: 600Gi
      volumeMode: Filesystem

Step 4: Modify "updateStrategy"

Change the "updateStrategy" type to "OnDelete" within the sts.yaml file:

  updateStrategy:
    type: OnDelete

Step 5: Update memphis-config ConfigMap

Update the memphis-config ConfigMap with the new size for the storageEngine space. Note that the reload feature will not work, so you will need to restart all the brokers one by one manually. Modify the storageEngine section as follows:

storageEngine {
  max_mem: 8Gi
  store_dir: /data

  max_file:600Gi
}

Step 6: Delete the Current StatefulSet with the Orphan Option

Execute the following command to delete the current StatefulSet, ensuring the use of the --cascade=orphan option:

kubectl delete statefulset --cascade=orphan memphis -n memphis

Step 7: Deploy Edited StatefulSet

Deploy the edited StatefulSet to the cluster using the sts.yaml file:

kubectl apply -f sts.yaml -n memphis

Step 8: Delete and Validate Pod Updates

Delete the Pods individually and confirm that they have the new configuration and retain the previous data.

Important!!! Deleting and validating stateful resources one by one is crucial for a smooth migration process.