Consistency groups

Consistency groups are used for controlling error situations and for maintaining data consistency at a recovery site.

You can create paths from a source LSS to a target LSS that have the consistency group option enabled. All volume pairs in remote mirror and copy relationships from this source to target LSS (sharing the same paths) belong to the consistency group. Other volumes pairs are not affected.

With the consistency group option enabled, when an error occurs on any volume pairs or on the links that are associated with these LSS pairs, an alert is issued and I/O to all duplex remote mirror and copy volumes on LSS pairs will be queued either until a consistency group created operation is run or the consistency group timeout time expires, allowing external automation to use the consistency group created operation to create a dependent write consistent set of target volumes over any number of LSS and disk storage systems.

External automation or command-line interface commands can issue a freeze to all LSS pairs that relate to the application, affecting the set of source volumes in the consistency group. The freeze operation causes the source volumes to suspend and their associated paths to end. Therefore, all associated updates at the recovery site are halted, resulting in a consistent data checkpoint. The automation program is not a part of the software provided with the storage system; it must be supplied by the user. However, IBM has offerings to assist with this automation. For more information, contact your IBM storage representative.

A consistency grouping timeout function gives automation software the time to issue the automated requests to ensure consistency. The consistency grouping timeout is a value that specifies the length of time that the volume remains in long-busy (unavailable) following an error. The default value for this parameter is set to two minutes.

A consistency group volume pair that is suspended will not go into an extended long busy state unless the primary volume is in a full duplex state before the suspension.

Note: The consistency group option is not required for Global Mirror or for Global Copy. The Global Copy target volumes at the recovery site are likely to be inconsistent. Consistency of the volume pairs can be ensured only when all updates are transmitted and copied to the recovery site while the application write operations are quiesced. Global Mirror uses a function called a session to control the formation of consistency groups. (A session is a collection of volumes across multiple storage systems that are managed together to create consistent copies of data.) The formation of these consistency groups is coordinated by the "master" storage system, which sends commands over remote mirror and copy links to its "subordinate" storage systems.

The following describes the behavior of these volume pairs when an error occurs that affects a volume in a consistency group:

Without the consistency group option enabled, the storage system causes the volume where the error is detected to enter a suspended state, but updates to that volume are still allowed.
With the consistency group option enabled, the volume becomes suspended and additionally enters a long-busy state. Updates are not possible to volumes from the source LSS to target LSS that are sharing the same path.

A critical mode determines the behavior of the remote mirror and copy pairs or consistency groups after a failure in communication between the source and target storage systems when all the paths between a pair are lost.