12.2 Guidelines for Using Pool Cluster Resources

Consider the guidelines in this section when working with shared NSS pools and volumes in the cluster:

12.2.1 Guidelines for Shared Pools

  • When the pool cluster resource is brought online, the pool is automatically activated by the resource load script. You typically do not activate the pool at the terminal console.

  • The unit of failover is the device (disk or LUN) that contains the pool. If you use multiple devices in the pool, all of those devices must be failed over together.

    IMPORTANT:As a best practice, we recommend a single LUN (or device) per shared pool, and one volume per pool. (one LUN=>one pool=>one volume)

  • If you delete a pool cluster resource, OES Cluster Services automatically removes the Pool Resource object and the virtual server object from eDirectory. After you delete the resource, you can unshare the device and use the pool locally, or you can delete the pool and its volumes.

    Ensure that you offline the cluster resource before you attempt to delete the cluster resource or the pool. See Section 10.16, Deleting Cluster Resources, or Disabling Clustering for a Pool, LVM Volume Group, or Service.

  • Before you rename a cluster-enabled pool, ensure that you offline the pool resource, activate the pool by using iManager or NSSMU, rename the pool, deactivate the pool, then online the pool resource.

    OES Cluster Services automatically updates the pool resource load and unload scripts to reflect the name change. Also, NSS automatically changes the Pool Resource object name in eDirectory.

12.2.2 Guidelines for Shared Volumes

You must create at least one volume on a shared pool before you migrate it. As a best practice, we recommend that you create only one volume per shared pool.

When you create a volume, commands are added to the pool resource load and unload scripts to automatically mount and dismount the volume when the scripts run. You can modify the load script to comment out the mount command so that you can manually mount the volume on a node in the cluster where the pool resource has been activated.

Volumes on cluster-enabled pools do not appear as independent cluster resources. A cluster-enabled volume does not have an associated load and unload script or an assigned IP address. If you want each volume to be in a separate cluster resource, each volume must have its own pool.

When a server fails, the cluster resources fail over to other servers in the cluster. Because the cluster-enabled pool fails over, all volumes in the pool also fail over, but only the volumes that have been added to the load script are automatically mounted. Any volumes in the pool that are not in the load script must be mounted manually. For this reason, volumes that you do not want to fail over should be in separate pools that are not cluster-enabled.

When you create an encrypted NSS volume in a shared pool, you must mount the volume manually by using NSSMU and enter the password. NSS uses the password to create a key. Instead of storing it in the server memory as it does for non-shared volumes, NSS asks OES Cluster Services to store the key and to pass it to the other nodes. After all servers hold the key, the volume is available for access as long as any one of the servers is still participating actively in the cluster. If all of the servers in the cluster fail, you must repeat this manual mounting procedure when you recover the cluster and restart services.

If you delete a volume from a cluster-enabled pool, OES Cluster Services automatically removes the volume mount command from the pool cluster resource load script.