Scheduler Examples

Scheduling modes

There are two modes available to enforce pod locality for Ondat Volumes. The mode is defined by labeling the Pods.

Labels

Locality Mode	Label	Description
Preferred (Default)	`storageos.com/locality: preferred`	Best effort placement of the Pod along its Ondat Volume
Strict	`storageos.com/locality: strict`	Enforce placement of the Pod alongside the Primary Volume or do not schedule

Preferred (Default)

The Pod will be placed along with its data if possible. Otherwise, it will be placed along the volume replicas. If both scenarios are not possible, the Pod will start on a different node and Ondat will grant accessibility to the data via network.

You can explicitly guarantee that mode by setting the label storageos.com/locality: "preferred"to the Pod.

For instance a Pod would set the labels as:

apiVersion: v1
kind: Pod
metadata:
  name: pod
  labels:
    storageos.com/locality: "preferred"
spec:
  containers:
    - name: debian
      image: debian:9-slim
      command: ["/bin/sleep"]
      args: [ "3600" ]
      volumeMounts:
        - mountPath: /mnt
          name: v1
  volumes:
    - name: v1
      persistentVolumeClaim:
        claimName: persistent-volume # ----> Ondat PVC

Strict

The Pod will be enforced to be placed along its data. Hence, in the same node as the Primary Volume. In case of that not being possible, the Pod will remain in pending state until that premise is fulfilled.

You can enable this mode by setting the label storageos.com/locality: "strict" to the Pod.

For instance an StatefulSet would set the labels as:

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: mysql
spec:
  selector:
    matchLabels:
      app: mysql
      storageos.com/locality: "strict"   # Matching label to the template
  serviceName: mysql
  replicas: 1
  template:
    metadata:
      labels:
        app: mysql
        storageos.com/locality: "strict" # Inherited label by the Pod
    spec:
      serviceAccountName: mysql
      containers:
      - name: mysql
        image: mysql:5.7
...
  volumeClaimTemplates:
  - metadata:
      name: data
      labels:
        app: mysql
    spec:
      accessModes: ["ReadWriteOnce"]
      storageClassName: "storageos" # StorageClass for your cluster
      resources:
        requests:
          storage: 50Gi

Explicit SchedulerName

It is not necessary to explicitly set the SchedulerName as the Admission Controller automatically populates the PodSpec field. Set the SchedulerName in your manifests, manually, only if you disable or can’t execute the Ondat Admission Controller.

Kubernetes allows the use of different schedulers by setting the field .spec.schedulerName: storageos-scheduler.

For instance a Pod manifest utilising the Ondat scheduler would appear as follows:

apiVersion: v1
kind: Pod
metadata:
  name: d1
spec:
  schedulerName: storageos-scheduler # --> Ondat Scheduler
                                     # No need if using Admission Controller
                                     # (enabled by default)
  containers:
    - name: debian
      image: debian:9-slim
      command: ["/bin/sleep"]
      args: [ "3600" ]
      volumeMounts:
        - mountPath: /mnt
          name: v1
  volumes:
    - name: v1
      persistentVolumeClaim:
        claimName: persistent-volume # ----> Ondat PVC