Mount two persistent volume claims in one deployment cause error

12/14/2018

I created two PersistentVolumeClaims(one for redis, one for persistent logs) and tried to mount both in a single deployment, but after creating the deployment, I get the following error:

nodes are available: 3 node(s) didn't match node selector, 4 node(s) had no available volume zone.

However as soon as I remove one PVC from the deployment yaml file, it works fine. I am running it on Google Cloud Platform using Kubernetes engine.

pvc1:

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: pvc-log
spec:
  accessModes:
    - ReadWriteOnce
  volumeMode: Filesystem
  resources:
    requests:
      storage: 20Gi
  storageClassName: standard

pvc2:

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: pvc-redis
spec:
  accessModes:
    - ReadWriteOnce
volumeMode: Filesystem
resources:
  requests:
    storage: 20Gi
storageClassName: standard

deployment:

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: 'prod-deployment'
spec:
  replicas: 1
template:
  metadata:
    labels:
      app: foo
      release: canary
      environment: production
  spec:
    containers:
      - name: api-server
        image: 'foo:latest'
        volumeMounts:
          - mountPath: /logs
            name: log-storage
      - name: redis
        image: 'redis'
        volumeMounts:
          - mountPath: /data
            name: redis-data
    volumes:
      - name: redis-data
        persistentVolumeClaim:
          claimName: pvc-redis
      - name: log-storage
        persistentVolumeClaim:
          claimName: pvc-log
-- Zhiliang Xing
google-kubernetes-engine
kubernetes

2 Answers

12/15/2018

It is rejected by scheduler for "NoVolumeZoneConflict" predicate, here is the declaration: https://github.com/kubernetes/kubernetes/blob/master/pkg/scheduler/algorithm/predicates/predicates.go#L564 (since I do not find any better doc, but I think the comment in code is really clear for this).

And as Rico said, you have to limit the volume zone for volumes in a pod via maybe storageclass or pv directly but not recommended.

-- c0rehe110
Source: StackOverflow

12/15/2018

This is similar to this. It's most likely due to a PVC trying to create a volume on an availability zone where you don't have a node in. You can try restricting the standard StorageClass to just the availability zones where you have Kubernetes nodes. Something like this:

kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: standard
provisioner: kubernetes.io/gce-pd
parameters:
  type: pd-standard
allowedTopologies:
- matchLabelExpressions:
  - key: failure-domain.beta.kubernetes.io/zone
    values:
    - us-central1-a
    - us-central1-b
-- Rico
Source: StackOverflow