Building Disaster Recovery Serviceguard Solutions Using Metrocluster with EMC SRDF

Table 6 Package startup behavior in various failure scenarios (continued)
Metrocluster
behaviourAUTO parameters
SRDF
StatesFailover/Failback
the package
directory and restart
the package. To
automate failover,
set AUTOSPLITR1 to
1.
By default, the
package fails to
AUTOR2XXNLPartitioned+SuspendedWDMIXED_STATEsync or
async
Failover to the
recovery site
start in this(R2) when the
condition. To startSRDF Links
the package, createare in mixed
a FORCEFLAG instate. ( This
the packagecan happen
directory and restartwith
the package. Toconsistency
automate failover,groups where
set AUTOR2XXNL to
1.
one link is in
Partitioned
state and the
other is in
Suspended
state).
Package Failover using SRDF/Asynchronous
The EMC Solutions Enabler provides a control operation checkpoint to confirm that the data written
in the current SRDF/Asynchronous cycle has been successfully committed to the R2 side. When a
package fails over to secondary site, Metrocluster with EMC SRDF ensures the most current data
when the SRDF link is still up. Metrocluster with EMC SRDF invokes the action checkpoint prior to
failover to the storage. Since the checkpoint operation prolongs the failover time for a package to
start, the duration for a package to start on R2 side is longer. The time taken to complete the
checkpoint operation depends on the cycle time configured which determines the amount of data
outstanding on the R1 site.
Failure scenarios in a complex workload
This section elaborates the failure scenarios that might occur when a complex workload is configured
using Site Aware Disaster Tolerant Architecture (SADTA).
Site failover
When the Site Controller package discovers that a running package configuration of a disaster
recovery complex workload has failed in the Metrocluster, or that the site hosting it has failed, it
fails over to the remote site node and initiates a site failover from the remote node. The site failover
starts the adoptive complex-workload package configuration by starting the packages configured
on the remote site.
The Site Controller package monitors the active complex-workload packages, according to the
configuration, to detect a failure and initiate a site failover. When the complex-workload packages
are configured using the critical_package attribute, the Site Controller package detects and
initiates a site failover even if one of the critical packages fail. In a configuration where all the
packages in the complex workload are configured with the managed_package attribute, the Site
Controller package detects a failure and initiates site failover based on the cumulative status of all
the configured managed packages.
A complex-workload package that has failed or is halted, in addition to displaying a down state,
also displays a halted status. A special flag, package_halted is set to no when the
Package Failover using SRDF/Asynchronous 57