I'm deploying a large Kube cluster where each node has 80 SSDs. Each disk is mounted as /data/diskN (disk0 - disk79) on each node.
Example PV:
apiVersion: v1
kind: PersistentVolume
metadata:
name: vol0022
spec:
accessModes:
- ReadWriteOnce
capacity:
storage: 100Gi
local:
path: /data/disk22
nodeAffinity:
required:
nodeSelectorTerms:
- matchExpressions:
- key: "storage"
operator: In
values:
- "true"
persistentVolumeReclaimPolicy: Retain
volumeMode: FilesystemDo I merely create enough PV's to cover (num_nodes * num_disk_per_node)?
ie:
If I have 10 nodes, I'll have 10 PV's where the path is "/data/disk0", another 10 where the path is "/data/disk1", and so on.
Thx
Yes. If you try to create Persistent Volume for each of them.
A better way is to use a "StorageClass" (more information) like "Glusterfs" which will create a cluster for all the disk. Then you can directly create a PersistentVolumeClaim to claim some space without creating a PersistentVolume.