Adding local scratch devices to DataStage pods

This example assumes that you already attached a new device to the worker node and its device is /dev/sdb. Log in to the worker node with the core user. All operations in this step require root privilege, so use the sudo su - command to become the root user.

[core@worker1 ~]$ sudo su -

1. Create the file system on the new device and mount it at /var/mnt/scratch:

   mkfs.ext4 /dev/sdb
   mkdir /var/mnt/scratch
   mount /dev/sdb /var/mnt/scratch
   chmod -R 777 /var/mnt/scratch
   semanage fcontext -a -t svirt_sandbox_file_t '/var/mnt/scratch(/.*)?'
   restorecon -RFv /var/mnt/scratch

   This set of commands also assigns the svirt_sandbox_file_t label to the directory to allow the access from the container, which is necessary because SELinux is turned on by default on CoreOS.

2. Make the mount volume persistent.
   1. Create a mount file in the /etc/systemd/system directory. The naming of the file is important. You need to name the file like the following example:

      <directory_path>-<mount_name>.mount
      

   2. Put a dash "-" between the directories. In this example, the volume is mounted at /var/mnt/scratch. The file name must be var-mnt-scratch.mount. If you do not follow this naming rule, the command to enable this mount fails. See the following example content:

      [Unit]
      Description = Mount a device for scratch
      
      [Mount]
      What=/dev/sdb
      Where=/var/mnt/scratch
      Type=ext4
      
      [Install]
      WantedBy = local-fs.target

   3. Run the systemctl command:

      [root@worker1 ~]# systemctl enable --now /etc/systemd/system/var-mnt-scratch.mount
      

   4. Repeat these steps on all nodes.
      Tip: If you create a script, this process can be much easier.

1. Log off from the worker node and come back to your workstation. The device is available on the worker nodes and a PV can be created with it.
2. Create the PV according to the following example:

   cat << EOF | oc -n $CPD_INSTANCE_NS apply -f -
   apiVersion: v1
   kind: PersistentVolume
   metadata:
     name: pv-scratch
     labels:
       type: local
   spec:
     storageClassName: local-scratch-class
     capacity:
       storage: 100Gi
     accessModes:
       - ReadWriteOnce
     hostPath:
       path: "/var/mnt/scratch"
   EOF
   

3. Create a persistent volume claim (PVC) that is to be used in the DataStage engine pods:

   cat << EOF | oc -n $CPD_INSTANCE_NS apply -f -
   apiVersion: v1
   kind: PersistentVolumeClaim
   metadata:
     name: pvc-scratch-manual
   spec:
     storageClassName: local-scratch-class
     accessModes:
       - ReadWriteOnce
     resources:
       requests:
         storage: 100Gi
   EOF
   

4. Check to see whether the PV was created:

   [root@mymachine ~]# oc get pvc | grep scratch
   pvc-scratch-manual                            Bound    pv-scratch                                 100Gi       RWO            local-scratch-class   22d

   The status Bound appears if the process was successful.

To mount the new PVC to a PX runtime instance, add the PVC to the PX runtime custom resource (CR) under additional_storage.

1. Retrieve the PX runtime instances:

   oc -n $CPD_INSTANCE_NS get pxruntime
   

2. Edit the PX runtime instance:

   oc -n $CPD_INSTANCE_NS edit pxruntime <pxruntime-cr-name>
   

3. Add the new PVC to additional_storage:

   spec:
     additional_storage:                        # mount additional persistent volumes
     - mount_path: /scratch                     # the path to mount the persistent volume
       pvc_name: pvc-scratch-manual            # the name of the associated persistent volume claim
     description: The default DataStage runtime instance
     license:
       accept: true
     parameters:
       scaleConfig: small
       storageClass: nfs-client
       storageSize: 10Gi
   

You can create a new configuration file or modify the existing ones. Change the line that starts with resource scratchdisk as shown in the following example:

{
  node "conductor"
  {
   fastname "$conductor" 
   pools "conductor" 
   resource disk "/px-storage/pds_files/node1" {pool "" "export" "node1"}
   resource scratchdisk "/scratch" {pool ""}
  }
  node "compute"
  {
   fastname "$pod" 
   pools "" 
   resource disk "/px-storage/pds_files/node#" {pool "" "export" "node2"}
   resource scratchdisk "/scratch" {pool ""}
  }
}