CVM maintains connectivity to shared disk groups as long as at least one node in the cluster has access to the configuration copies. The master node performs changes to the configuration and propagates the changes to all of the slave nodes. If the master node loses access to the copies of the configuration, the master node sends the writes for the configuration change to a slave node that has access. The slave node implements the changes. This behavior enables the disk group to stay active, even when the master node loses access to the shared disk group configuration. If all nodes lose access to the shared disk group, the disk group is disabled.

In previous releases prior to release 6.0, you could configure a disk group failure policy for a shared disk group. The disk group failure policy is no longer supported for disk groups with the latest disk group version and cluster protocol version.

If the cluster protocol version is less than 110, or the disk group version is less than 170, the disk group failure policy (dgfailpolicy) determines the behavior after failure.

See the documentation from the relevant release of VxVM.

Disk group configurations are handled this way regardless of the setting for the ioship policy. The ioship policy controls redirection of application I/Os.

See About redirection of application I/Os with CVM I/O shipping.