HP StorageWorks SAN Virtualization Services Platform administrator guide (5697-0204, January 2010)

NOTE:
When using synchronous mirrors to protect multiple virtual disks that belong to the same application,
it is best practice to have all of the virtual disks active on the same DPM. This is especially true when
in the stretched domain, so as to have all virtual disks of that application fail over together or not at
all.
Site failures
In this topology, each site failure affects the surviving site. The recovery order, however, depends on
the site that fails. Always complete the synchronous mirror groups' recovery to other domains before
you recover the synchronous mirror groups between the sites of the failed domain. Recovering to other
domains first and to other sites second enables you to perform a full recovery using only Force Resume
procedures. If you do not follow the recommended recovery order, you will have to use Force Delete
procedures to perform a recovery.
Before you start the procedure, make sure that the site is down, as the site might be only disconnected.
You must restore the inter-switch link connection to recover the site if it is disconnected.
Recovering from a Site A failure
When Site A fails, you must recover the synchronous mirror groups that mirror to Site B. On the
surviving Site B, recover the mirrors with Site A. Follow the procedure in “Recovering synchronous
mirror groups on the surviving site” on page 107.
Recovering from a Site B failure
When Site B fails you must recover the synchronous mirror groups that mirror to Site A. On the surviving
Site A, recover the mirrors with Site B. Follow the procedure in “Recovering synchronous mirror groups
on the surviving site” on page 107.
Synchronous mirroring in a single stretched SVSP domain
Solution topology
Figure 35 shows the topology of a synchronous mirror configuration in a single SVSP domain with
two SVSP sites. Figure 35 shows two sites that represent two independent data centers. Each site is
organized in a redundant SAN configuration and includes:
• At least one storage array with redundant controllers
• Two fabrics
• One VSM server
• One Data Path Module (DPM) per DPM group
• Servers, with each server having at least two host bus adapters (HBAs)
The fabrics of the two sites are interconnected to form two “cross-site fabrics.” The cross-site fabrics
are indicated by the blue and the red lines in Figure 1.
Site failure recovery with synchronous mirroring104