Can I use a hostPath in the ModelOps file system for file-based Source and Sink Channels?

Use for testing only, not for Production.

Yes, hostPath can work, with limitations. Any hostPath volumes are exclusively used for single node clusters. If your pod moves to another node in a multi-node cluster, your hostPath storage doesn't move with it. The following configuration will not handle multi-node clusters or a node failure, so use this only for simple testing.

Instead, the recommended solutions for multi-node clusters are to use NFS mounts (for EKS and OpenShift), as described in the ModelOps documentation. For Azure use an Azure File Share.

STEPS TO USE hostPath VOLUME CONFIGURATION...

1. Run Lens. In the "Nodes" view identify the nodes running the cluster. Select the node to use for storage. You should be able to open a terminal there and you will be 'root'. Run the following commands from that terminal.

Note that you are only running Lens on Windows. It connects to your AKS Cluster and your new volume is in the Linux cluster, not on Windows. The following commands are to be run in the node that represents your storage volume (the PersistentVolume nodes created by the 'kubectl' examples that follow).

2. Create the File Source Volume

Source 10GB Persistent Volume

kubectl apply -f - <<!
apiVersion: v1
kind: PersistentVolume
metadata:
   name: task-pv-volume-source
   labels:
     type: local
     usage: task-pv-volume-source
   annotations:
     description: "file-datasource-local"
spec:
   capacity:
     storage: 10Gi
   accessModes:
     - ReadWriteMany
   hostPath:
     path: "/tmp/source"
!

Source 10GB Persistent Volume Claim

kubectl apply -f - <<!
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
   name: task-pv-claim
   namespace: datachannels
   annotations:
     description: "file-datasource-local"
     volume.beta.kubernetes.io/storage-class: ""
 spec:
   accessModes:
     - ReadWriteMany
   resources:
     requests:
       storage: 10Gi
   selector:
     matchLabels:
         usage: task-pv-volume-source
!

3. Create the File Sink Volume

Sink 10GB Persistent Volume

kubectl apply -f - <<!
apiVersion: v1
kind: PersistentVolume
metadata:
   name: task-pv-volume-sink
   labels:
     type: local
     usage: task-pv-volume-sink
   annotations:
     description: "file-datasink-local"
spec:
   capacity:
     storage: 10Gi
   accessModes:
     - ReadWriteMany
   hostPath:
     path: "/tmp/sink"
!

Sink 10GB Persistent Volume Claim

kubectl apply -f - <<!
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
   name: task-pv-claim-sink
   namespace: datachannels
   annotations:
     description: "file-datasink-local"
     volume.beta.kubernetes.io/storage-class: ""
 spec:
   accessModes:
     - ReadWriteMany
   resources:
     requests:
       storage: 10Gi
   selector:
     matchLabels:
         usage: task-pv-volume-sink
!

4. Open the directory permissions for write

chmod 777 /tmp/sink

After these commands are run you may create Source and Sink channels in your single ModelOps instance. Only use this configuration for temporary testing.