This applies to Spectrum CDM: Backup on AIX nodes managed by PowerHA

The High Availability configuration of Power AIX systems is made up of two or more real nodes to which volumes and IP addresses are associated and then managed by PowerHA.

In this configuration there are therefore at least three types of volumes:

- local volumes (example OS);

- database volumes (both on filesystem and on ASM disks);

- shared volumes managed by HA and therefore opened on the active node (for example containing data, application binaries or other).

For the first type of volume there is the "Filesystem" type: real nodes are created and the filesystems to be copied are associated;

For the second type of volume there is the "*Database*" type: the virtual node is created and the SID/s to be copied is selected;

For the third type of volumes, there is no node type. The only possibility is to associate the shared volumes to one of the real nodes.

By doing this, when the jobs are executed, the following problems occur:

1) Shared volumes can only be copied from the real node that owns them. This means that the copy will fail if the volume is moved to the secondary node;

2) The virtual node clearly leans on the real node. If the database inventory job runs after the filesystem inventory job, the active real node loses ownership of the filesystems (even local ones) and that node's copy fails;

3) When the inventory jobs are executed, if the "*Database*" one passes after the "Filesystem" one, the active node loses the configuration of the filesystems because they become the property of the "*Database*" node and the copy fails . Furthermore, from the database node it is not possible to save those volumes because they are not shown.

4) It is not possible to define more than one node with the same IP address (or hostname) and therefore it is not possible to associate the shared volumes to another virtual node.

The need would be to be able to have a new "Clustered Filesystem" type that allows associating shared volumes and whose inventory does not steal ownership from real nodes.

The inventory of the “*Database*” nodes should also do the same