HP Matrix Operating Environment 7.0 Recovery Management User Guide

NOTE: Matrix recovery management is able to prevent split-brain from occurring during an
unplanned failover, by regulating the auto-power configuration of managed nodes (whether
virtual or physical) that are assigned to DR Protected logical servers so that they do not
automatically power up after an outage. If, for example, a site loses power and site failover
is invoked, the site where the power outage occurred will not resume running the DR Protected
logical servers when power is restored. The managed nodes (whether VC blades or virtual
machines) assigned to DR Protected logical servers will stay powered down (and resources
are will remain unassigned) until an Activate... operation is invoked.
NOTE: When a Deactivate... operation is invoked, the status of Recovery Groups that are
being deactivated will not change on the Matrix recovery management Sites or Jobs screens,
until the Deactivate... operation has completed. While the Deactivate... operation is running,
there may be some Recovery Groups that have been deactivated while others are still in the
process of being deactivated.
NOTE: If the Matrix recovery management configuration has been changed since the failover
occurred (for example, a new Recovery Group was created), the sites must be brought into
sync by making appropriate configuration changes. The Matrix recovery management Site
configuration export and import operations can be used for this purpose.
NOTE: A successful activate or deactivate operation ensures that all of the Recovery Groups
within a Recovery Group Set are in the same state (enabled or disabled). However, certain
operations (for example, a Recovery Group edit to change site preference) may result in some
Recovery Groups within a Recovery Group Set being enabled and others being disabled. You
must run an activate or deactivate operation on a Recovery Group Set to ensure that all
Recovery Groups in that Recovery Group Set are in the same state (enabled or disabled).
Target selection and parallelism during an activation operation
The HP Matrix OE logical server management (logical server management) component in the HP
Matrix Operating Environment supports the concept of targets that are most suitable for activating
a logical server based on various criteria, for example, an application may need to run on VC
hosted logical servers only, to meet a performance requirement.
DR Protected logical servers that can run on both physical and virtual targets (cross technology
logical servers) will be placed on the target type specified as preferred in the site configuration (P
for physical or V for virtual), based on availability. If the preferred target type is not available,
Matrix recovery management will ignore the target type preference and place cross technology
logical servers on available supported targets.
Matrix OE logical server management allows logical servers to be activated in parallel, exploiting
parallelism that is available in the managed infrastructure. Matrix recovery management exploits
the parallelism of activation available from logical server management when performing an
Activate... operation, to reduce failover time. Two user accessible settings are available to influence
Matrix recovery management behavior in this area:
• If there are workloads that you want to bring up first during the failover process, the Recovery
Group Start Order values for the associated Recovery Groups can be set lower than the values
you set for workloads that can be brought up later in the failover process. Matrix recovery
management will ensure that logical servers in a Recovery Group with a lower Recovery Group
Start Order value will be activated before logical servers in a Recovery Group with a higher
Recovery Group Start Order value are activated.
• Use the Recovery Group Power Up Delay parameter, if you want to ensure that the logical
servers in a Recovery Group boot in a staggered fashion during the failover process. The
Recovery Group Power Up Delay parameter sets a minimum delay between the time when
one logical server in the Recovery Group begins its boot process and the time when the next
logical server in the Recovery Group begins its boot process.
24 Testing and failover operations