Create PX-Fast PVCs

PX-Fast is a Portworx feature that enables an accelerated IO path for the volumes that meet certain prerequisites. It is optimized for workloads requiring consistent low latencies. PX-Fast is built on top of a Portworx PX-StoreV2 datastore.

Prerequisites

You must have deployed Portworx with PX-StoreV2 datastore.

Create PX-Fast PVCs

Perform the following steps to create PX-Fast volumes.

Create a StorageClass

1. Save the following as a YAML file:

   kind: StorageClass
   apiVersion: storage.k8s.io/v1
   metadata:
       name: px-fast
   provisioner: kubernetes.io/portworx-volume
   parameters:
       repl: "1"
       fastpath: "true"
   allowVolumeExpansion: true

   Note that this StorageClass will have 1 replica, and the Portworx volumes referring to this StorageClass will also have 1 replica.

2. Run the following kubectl apply command to create a StorageClass:

   kubectl apply -f <px-fast>.yaml

   storageclass.storage.k8s.io/px-fast-sc created

3. Verify that the StorageClass is created:

   kubectl describe storageclass px-fast-sc

   Provisioner:           kubernetes.io/portworx-volume
   Parameters:            fastpath=true,repl=1
   AllowVolumeExpansion:  <unset>
   MountOptions:          <none>
   ReclaimPolicy:         Delete
   VolumeBindingMode:     Immediate
   Events:                <none>

Create a PVC

1. Save the following as YAML file:

   kind: PersistentVolumeClaim
   apiVersion: v1
   metadata:
       name: px-fast-pvc
   spec:
       storageClassName: px-fast
       accessModes:
           - ReadWriteOnce
       resources:
           requests:
               storage: 1Gi

2. Run the following kubectl apply command to create a PX-Fast PVC:

   kubectl apply -f <px-pvc>.yaml

3. Verify that a persistent volume claim is created:

   kubectl get pvc

Deploy an application using a PX-Fast PVC

To utilize PX-Fast on the above PVCs, the application pods must run on the same node where the volume replica exists. The following sample deployment spec uses the stork.libopenstorage.org/preferLocalNodeOnly: "true" annotation, which enforces the pod to be scheduled on the node where the volume replica exists:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: mysql
spec:
  selector:
    matchLabels:
      app: mysql
  strategy:
    rollingUpdate:
      maxSurge: 1
      maxUnavailable: 1
    type: RollingUpdate
  replicas: 1
  template:
    metadata:
      annotations:
        stork.libopenstorage.org/preferLocalNodeOnly: "true"
      labels:
        app: mysql
    spec:
      schedulerName: stork
      containers:
      - image: mysql:5.6
        name: mysql
        env:
        - name: MYSQL_ROOT_PASSWORD
          value: password
        ports:
        - containerPort: 3306
        volumeMounts:
        - name: mysql-persistent-storage
          mountPath: /var/lib/mysql
      volumes:
      - name: mysql-persistent-storage
        persistentVolumeClaim:
          claimName: px-fast-pvc

Note that the schedulerNamefield in the above spec is set to stork. From Stork version 2.9.0 or newer, the schedulerName is set to stork by default.

The annotation stork.libopenstorage.org/preferLocalNodeOnly: "true" enforces the pod to be scheduled on the node where the volume replica exists. To avoid this enforced behavior, you can choose not to specify the stork.libopenstorage.org/preferLocalNodeOnly: true annotation. In such a case, Stork will do best-effort to place the application pod on the node where the replica exists, but if the application pod gets attached to a non-replica node, it cannot use PX-Fast.

Reference

PX-Fast volumes using pxctl