Nodegroup Plugin User Guide

Introduction

Nodegroup plugin is designed to isolate resources by assigning labels to nodes and set node label affinty on Queue.

Usage

assign label to node

Assign label to node, label key is volcano.sh/nodegroup-name.

kubectl label nodes <nodename> volcano.sh/nodegroup-name=<groupname>

configure queue

Create queue and bind nodegroup to it.

apiVersion: scheduling.volcano.sh/v1beta1
kind: Queue
metadata:
  name: default
  spec:
    reclaimable: true
    weight: 1
    affinity:            # added field
      nodeGroupAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
        - <groupname>
        preferredDuringSchedulingIgnoredDuringExecution:
        - <groupname>
      nodeGroupAntiAffinity:
        requiredDuringSchedulingIgnoredDuringExecution:
        - <groupname>
        preferredDuringSchedulingIgnoredDuringExecution:
        - <groupname>

submit a vcjob

submit vcjob job-1 to default queue.

$ cat <<EOF | kubectl apply -f -
apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: job-1
spec:
  minAvailable: 1
  schedulerName: volcano
  queue: default
  policies:
    - event: PodEvicted
      action: RestartJob
  tasks:
    - replicas: 1
      name: nginx
      policies:
      - event: TaskCompleted
        action: CompleteJob
      template:
        spec:
          containers:
            - command:
              - sleep
              - 10m
              image: nginx:latest
              name: nginx
              resources:
                requests:
                  cpu: 1
                limits:
                  cpu: 1
          restartPolicy: Never
EOF

validate queue affinity and antiAffinity rules is effected

Query pod information and verify whether the pod has been scheduled on the correct node. The pod should be scheduled on nodes with label nodeGroupAffinity.requiredDuringSchedulingIgnoredDuringExecution or nodeGroupAffinity.preferredDuringSchedulingIgnoredDuringExecution. If not, the pod should be scheduled on nodes with label of nodeGroupAntiAffinity.preferredDuringSchedulingIgnoredDuringExecution. Specifically, the pod must not be scheduled on nodes with the label nodeGroupAntiAffinity.requiredDuringSchedulingIgnoredDuringExecution.

kubectl get po job-1-nginx-0 -o wide 

How the Nodegroup Plugin Works

The nodegroup design document provides the most detailed information about the node group. There are some tips to help avoid certain issues.These tips are based on a four-nodes cluster and vcjob called job-1:

Node	Label
node1	groupname1
node2	groupname2
node3	groupname3
node4	groupname4

apiVersion: batch.volcano.sh/v1alpha1
kind: Job
metadata:
  name: job-1
spec:
  minAvailable: 1
  schedulerName: volcano
  queue: default
  policies:
    - event: PodEvicted
      action: RestartJob
  tasks:
    - replicas: 1
      name: nginx
      policies:
      - event: TaskCompleted
        action: CompleteJob
      template:
        spec:
          containers:
            - command:
              - sleep
              - 10m
              image: nginx:latest
              name: nginx
              resources:
                requests:
                  cpu: 1
                limits:
                  cpu: 1
          restartPolicy: Never

1. Soft constraints are a subset of hard constraints, including both affinity and anti-affinity. Consider a queue setup as follows:

   apiVersion: scheduling.volcano.sh/v1beta1
   kind: Queue
   metadata:
     name: default
     spec:
       reclaimable: true
       weight: 1
       affinity:            # added field
         nodeGroupAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname1
           - gropuname2
           preferredDuringSchedulingIgnoredDuringExecution:
           - groupname1
         nodeGroupAntiAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname3
           - gropuname4
           preferredDuringSchedulingIgnoredDuringExecution:
           - groupname3
   

   This implies that tasks in the "default" queue will be scheduled on "groupname1" and "groupname2", with a preference for "groupname1" to run first. Tasks are restricted from running on "groupname3" and "groupname4". However, if the resources in other node groups are insufficient, the task can run on "nodegroup3".

2. If soft constraints do not form a subset of hard constraints, the queue configuration is incorrect, leading to tasks running on "groupname2":

   apiVersion: scheduling.volcano.sh/v1beta1
   kind: Queue
   metadata:
    name: default
    spec:
      reclaimable: true
      weight: 1
      affinity:            # added field
        nodeGroupAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
          - gropuname2
          preferredDuringSchedulingIgnoredDuringExecution:
          - groupname1
        nodeGroupAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
          - gropuname4
          preferredDuringSchedulingIgnoredDuringExecution:
          - groupname3
   

3. If there is a conflict between nodeGroupAffinity and nodeGroupAntiAffinity, nodeGroupAntiAffinity takes higher priority.

   apiVersion: scheduling.volcano.sh/v1beta1
   kind: Queue
   metadata:
     name: default
     spec:
       reclaimable: true
       weight: 1
       affinity:            # added field
         nodeGroupAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname1
           - gropuname2
           preferredDuringSchedulingIgnoredDuringExecution:
           - groupname1
         nodeGroupAntiAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname1
           preferredDuringSchedulingIgnoredDuringExecution:
           - gropuname2
   

   This implies that tasks in the "default" queue can only run on "groupname2".

4. Generally, tasks run on "groupname1" first because it is a soft constraint. However, the scoring function comprises several plugins, so the task may sometimes run on "groupname2".

   apiVersion: scheduling.volcano.sh/v1beta1
   kind: Queue
   metadata:
     name: default
     spec:
       reclaimable: true
       weight: 1
       affinity:            # added field
         nodeGroupAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname1
           - gropuname2
           preferredDuringSchedulingIgnoredDuringExecution:
           - groupname1
         nodeGroupAntiAffinity:
           requiredDuringSchedulingIgnoredDuringExecution:
           - groupname3
           - gropuname4
           preferredDuringSchedulingIgnoredDuringExecution:
           - groupname3