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

The end result is the same: the surviving site loses connectivity with the VSM server, the DPM,
application servers, and the storage arrays at the other site.
Depending on which reason leads to lost connectivity between the two sites, you can choose one of
these recovery options:
• A short outage is anticipated: If a site experiences a power outage or if the cables between the
two sites are disconnected, perform one of these actions before using the data on the surviving
site to bring the applications online, then:
• Restore power to the site that lost power.
• Reconnect any cables that might have become disconnected.
• A long outage is anticipated: If a site is completely destroyed, or the intersite connection lost, your
only option is to recover by using the data from the surviving site.
VSM does not automatically recover from the surviving site. You must first identify the reason for
failure, and then use the VSM synchronous mirror features to perform the recovery operation.
Synchronous mirror behavior as a result of a site failure or a site disconnection
• The active VSM remains active. The active VSM marks the synchronous mirror tasks of the setup
virtual disk to the other site as Failed. The status of the setup virtual disk synchronous mirror group
changes to Journal.
• The passive VSM tries to become active but fails. The Recovery tab on the VSM monitor indicates
a problem with the synchronous mirroring of the setup virtual disk. It happens because the VSM
cannot access the synchronous mirror copies of the setup virtual disk at the other site. The passive
VSM waits for your instructions on how to recover. Meanwhile, VSM alternates between the
Passive state and the Stopped state.
• Each DPM remains active for the synchronous mirror groups that it serviced before the site discon-
nection.
• Each DPM loses the access to the synchronous mirror storage on the other site. The active DPM
moves the relevant tasks to Failed status and the group to Partial status or Journal status, depending
on mirror mode (Synchronous Always or Automatic Failover).
• If the host that uses the synchronous mirror group is physically located on the same site as the
active DPM, and the synchronous mirror group mode is Automatic Failover, the host I/O operations
continue.
• If the host that used the synchronous mirror group lost connection with the active DPM, it sends a
failover request to the passive DPM.
• If the passive DPM and the active VSM reside on the same site, VSM approves the failover request
and appoints the passive DPM as the new active DPM for that synchronous mirror group. (VSM
approves the failover request because it sees the other DPM as absent.) The DPM then tries to start
the mirror and fails because it cannot access one of the tasks. You must direct the mirror on how
to recover. The synchronous mirror group status is Partial.
• If the passive DPM and the active VSM are not on the same site, the failover request fails, and the
host I/O operations fail. Manual intervention for recovery is expected.
Synchronous mirror failure analysis
1. Find out the actual status of the sites by checking the status of the individual components locally
on each site.
2. If the individual components are online, but the active VSM does not see those of the other site,
then the sites are probably just disconnected. Verify this is the case.
Site failure recovery with synchronous mirroring106