Concurrent access to volume groups
In concurrent access, the volume groups residing on shared disks are activated and can be simultaneously accessed in read/write mode by all active nodes in the cluster.
Accessing disks in concurrent mode requires that software packages running above the LVM coordinate their operations across the cluster to ensure data integrity. The AIX Journaled File System is not supported on volume groups accessed in concurrent mode. Consult the provider of any middleware to determine whether that middleware can reliably function in concurrent mode.
For concurrent access volume groups, PowerHA® SystemMirror® must ensure that no disk reserve is placed on the disk so that access by other nodes is possible.
The question arises as to whether any disk can be used in concurrent mode. The answer is no, because of a practical requirement to mirror data in a PowerHA SystemMirror cluster, so that data is not be lost if any of the disks fails. Even though most disks could be accessed without establishing a disk reserve on them, many disks still cannot be used in concurrent mode, because the LVM cannot simultaneously update all the copies of a single logical partition. Even when the write operationon a disk are done in parallel, they might not actually be completed at the same time, due to other activity on the disk or due to the physical characteristics of the disks. If one node writes data and the other node reads it, there is no guarantee that the reader gets the latest copy. Additionally, if the LVM cannot update its structures on a disk, it ceases to write to that disk and marks it as failed. In a multi-node cluster, this situation creates a single point of failure.