When datasets extend across multiple volumes, FlashCopy® must preserve the correct copy order to avoid inconsistent results. Consistency groups provide this protection by freezing I/O across all related source volumes while FlashCopy® pairs are being created. Once all relationships are established, the freeze is lifted and the snapshot is consistent across the entire data set.
The following figures show examples of using consistency groups:
- #1 and #4
- #2 and #5
- #3 and #6,
The figure, illustrates cross volume dependencies:
- Data 3 depends on Data 2
- Data 2 depends on Data 1
Without a consistency group, any writes occurring during pair creation can break dependency order. Consistency groups maintain correct write ordering by freezing I/O, creating all pairs, setting the bitmaps, then lifting the freeze. This ensures the target volumes receive a consistent data image.
The figure shows Data 2 in Volume #2 reference Data 1 in volume #1, and Data 3 in volume #3 references Data 2 in volume #2. Without consistency groups, when you try to copy Data 1 - 3 to volume #4 - #6, if Data 1 is updated before the copy operation of Data 2 is complete, the data consistency of copied Data 1 and copied Data 2 is not maintained.
Consistency groups maintain data consistency across volumes when they are copied using FlashCopy®. In the figure, FlashCopy® regards volumes #1 - #3 as a single consistency group, and suspends write operations from the host to the source volumes until the three relationships (Relationship A, Relationship B, and Relationship C) are created and all copy processing is finished. To suspend write operations to source volumes, you issue the command that creates the relationships with the appropriate "freeze" option set for your interface.
When FlashCopy® and TrueCopy for Mainframe (TCz) share volumes, the results of operations such as creating pairs and relationships will differ depending on whether you are using a consistency group.