With /FAILOVER-STORAGE-CLUSTER
you can switch the active primary storage system to the secondary storage system without interruptions (failover) during ongoing operation, e.g. for scheduled maintenance processes in the primary storage system. The BS2000 applications are continued without interruptions. For the BS2000, they are continued transparently on the same pubset and the same units, even if the storage system and the underlying volumes have changed due to the failover.
Therefore the direction of the internal replication changes, i.e. the volumes of the secondary storage system, are, as far as possible, mirrored to the volumes of the primary storage system with REC. This internal replication is not visible for SHC-OSD.
After a successful failover, the secondary storage system again is the active storage system. The primary storage system becomes the standby system.
The operating status of the Storage Cluster will then be FAILED-OVER
or FAILBACK-READY
.
A manual failover is only possible if the status is NORMAL
and the phase is NORMAL
. For example, if one or more volumes from the storage cluster are used as target units for an additive remote replication, the phase is COPYING
and a manual failover is rejected.
After a failure of the primary storage system, you can switch to the standby secondary storage system (failover) with /FAILOVER-STORAGE-CLUSTER ...,FORCE=YES
, if no automatic failover is set. The FORCE=YES
is only possible in case of a failure; in the NORMAL
operating status it is rejected.
The data on the secondary storage system are consistent with the data on the primary storage system at the time of failure, if the Storage Cluster was in NORMAL
operating mode prior to the failure.
Before a planed, manual failover, the configuration of the SPLIT-MODE
for the storage cluster should be checked. A failover with SPLIT-MODE=*READ
could lead to an interruption of the Application, due to the stop of IO activity.
To assure the availability of the Application, the SPLIT-MODE
should be *WRITE
bevore a manual failover.