Serviceguard Extended Distance Cluster (EDC) with VxVM/CVM Mirroring on HP-UX, May 2008

Figure 7 shows an example of the vxassist command with all of the recommended options for

creating an EDC volume—including the mirror configuration.

Data Change Object (DCO) volume for Dirty Region Log (DRL) and

FastMirrorResync (FMR)

When large mirrored volumes get out of sync due to a failure condition like system crash or

storage/link failure, or due to a maintenance operation, the resynchronization can take a long time.

VxVM/CVM includes the following two features to help in these circumstances:

• A Dirty Region Log (DRL) keeps track of the most recent writes and shortens the resynchronization

after a node failure. The DRL is used to recover after a node failure by synchronizing the regions of

a volume to which the failing node could have in-flight I/Os. This synchronization is done

automatically.

• A FastMirrorResync (FMR) map keeps track of changes to a volume while a plex is detached, due to

an I/O error, or “snapped off” for administrative purposes (for example, for making a backup).

After the storage link becomes available again, or when the snapped off volume is ready to be re-

attached to the original volume, the information in the FMR map is used to identify the regions that

need to be resynchronized. In the case of a link recovery, this resynchronization has to be started

manually. The longer the application sends I/Os to only one plex of the volume, the greater the

amount of data that will need to be resynchronized. To start the resynchronization as soon as

possible after the link becomes available again, the syslog should be monitored for events that

detached a plex.

Both DRL and FMR are implemented as bitmaps and are stored together in a version 20 DCO volume

that is attached to the “data” volume. Version 20 DCO has been available since VxVM version 4.0.

Storing DRL and FMR maps in the same VxVM object provides the following advantages:

• Reduced complexity by combining the individual VxVM objects into one, resulting in fewer objects

to manage

• Optimized updates of the DRL and FMR maps by updating both of them with one I/O, resulting in

less I/O overhead on writes