Compaq StorageWorks Data Replication Manager HSG80 ACS Version 8.5P Operations Guide

Managing Site Failover and Failback Procedures 5–3
Site Failover
If the initiator site is no longer available, or if there is anticipated down time that will
prevent operation there, you must decide whether or not to perform a site failover.
Performing a failover will enable the target site to assume the role of the initiator and
access data until the problem is resolved. Transferring control of system operation to the
target site ensures that there will be minimal interruption in data access after a failure.
NOTE: If you decide to perform a Failover operation, keep in mind that all components must be
failed over. Therefore, if only one component fails, fixing that single component may be
preferable to performing a complete failover. Also, it is important to verify that all components at
the target site are operational before you begin the site failover.
Table 5–1 outlines example scenarios that may call for a failover and those that may not.
NOTE: If one host in a multi-host environment fails, you must decide whether or not a failover is
the best course of action.
When you decide that a site failover is necessary, identify which scenario best describes
your situation: planned or unplanned failover.
The planned failover procedure will be used when failover is a scheduled event.
Otherwise, an unplanned failover will be used.
CAUTION: Be sure to follow the steps outlined in the section “Planned Failover and
Failback” accurately and completely, or you may incur data loss and extended
downtime.
Table 5–1 Failover Scenarios
When to Failover When Not to Failover (recommended)
■
Both controllers fail ■ Single failed switch
■ Extended power outage at the initiator site ■ Single fiber optic cable malfunctions
■ Both host adapters fail (non-clustered hosts) ■ Single controller fails
■ Both initiator switches fail ■ Single storageset fails
■ Disaster (flooding, fire, earthquake, terrorism,
etc.) that disables access to the subsystems
■ Single disk in redundant storageset fails
■ Scheduled event that will prevent computing
from the initiator site for an extended period
■ Target not in normal state
■ All hosts fail