Will local-path-provisioner separate pvc replicas across the cluster

[sadekunle]

I'm deploying a statefulset workload with 3 replicas. I see that the local-path-provisioner deploys all pvc replicas to one node. Is there a way to separate the pvc for each replica to a different node?

[yasker]

The scheduling of the storage is not controlled by the local path provisioner. It will waitForFirstConsumer (a.k.a. pod creation/scheduling). So if you use anti-affinity for the pods during the first creation, the volumes should follow the pod's scheduling to the different nodes.

After first time, the PV would bound to a certain node and cannot be moved after.

[sadekunle]

Would I need to create a pvc for each replica? Trying to deploy the 3 replicas and after the first replica is deployed I get the following error.

Warning FailedScheduling default-scheduler 0/4 nodes are available: 1 node(s) didn't match pod affinity/anti-affinity, 1 node(s) didn't satisfy existing pods anti-affinity rules, 3 node(s) had volume node affinity conflict.

[yasker]

If there is only one PVC, it will only able to be on one node, since the data is stored in one node's hard drive.

[ThinCats]

Statefulset and volumeClaimTemplates may fit your needs

see https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/

[dfoxg]

I pointed my local-path-provisioner-Path to an glusterfs-mount, so all of my files are available on all k8s-nodes. How can I configure the local-path-provider that the error had volume node affinity conflict does not appear?