HP StorageWorks Storage Mirroring for Linux User's Guide (T2558-96078, February 2008)

8 - 1
8 X Replication
Replication
A file change, also called a file request or operation, is the creation, modification, or deletion of a file. This also includes file
attribute and permission changes. Replication is the real-time transmission of these file changes as they occur. The file changes
are replicated from the source to the target, maintaining an identical copy of data on the target machine.
Replication is the key to Storage Mirroring; without having identical copies of data on the source and target machines at all
times, the target would not be able to meet the high availability and disaster recovery needs of your complex network
environment.
Replication Options
Replication can be started automatically on connection, which is the default and recommended configuration, and it can also
be started and stopped manually at any time. Stopping replication is separate from stopping the transmission of data to the
target. When replication stops, data is not added to the queue on the source.
You can also turn off replication when you are not concerned about real-time backup, but want to use Storage Mirroring’s
block checksum mirror to periodically synchronize files.
Recovery From Network Losses
Storage Mirroring handles packet losses on the network as well. Storage Mirroring verifies the incoming sequencing before
looking at execution ordering. Here is an example.
1. The source sends packets 1, 2, 3, and 4 to the target.
2. The target receives packet 1. It expects 1, so it passes the operation to the holding/execution handlers and sends its
acknowledgement of packet 1 back to the source.
3. Packet 2 is lost on the network.
4. The target receives packet 3. It expects 2, so it discards packet 3 and resends its acknowledgement of operation 1 to
the source because it was the last successful operation.
5. Because the source already received an acknowledgement for packet 1, it puts any operations that are in the
acknowledgement queue (starting with operation 2) back on the front of the transmit queue to be sent again, in order.
Task Command Processing
Task command processing is a Storage Mirroring feature that allows you to insert and run tasks at various points during the
replication of data. Because the tasks are user-defined, you can achieve a wide variety of goals with this feature. For example,
you might insert a task to create a snapshot or backup on the target after a certain segment of data from the source has been
applied on the target.
Task command processing can be enabled from both the Management Console and the text clients but can only be initiated
using the DTCL
queuetask command through the text clients.
WARNING: When stopping replication, source/target data integrity is no longer guaranteed. A remirror will be necessary
to guarantee that the source and target data are identical.