HP Capacity Advisor 7.2 User Guide

This process affects only the system workload for the virtual machine. If monitored workloads are
defined inside the virtual machine, their data is tracked correctly even when the virtual machine
fails over to a new host. It is a good idea to create a workload inside any virtual machine that you
intend to fail over using Serviceguard to ensure that the utilization data of the virtual machine is
kept whole.
TIP:
Prevent data loss
To prevent the loss of utilization data when migrating Integrity VM guests using Serviceguard,
create a monitored workload for the processes of interest on the virtual machine.
If there is no monitored workload, then utilization data for the failed-over system is split across two
or more workloads. To prevent this, use the following procedure to treat these multiple workloads
as a single workload:
1. Create a scenario and include the current virtual machine host.
2. Within that scenario, create a new workload and import data from the unrepresented workload.
(Create two or three workloads, if needed.)
3. Place these new workloads in the current virtual machine.
4. When you edit or move one of these workloads, do the same with the others.
The virtual machine containing the multiple workloads can be moved within the scenario and can
give a complete picture of past utilization. The collection of these multiple workloads gives you a
nearly uninterrupted view of the historical utilization for the virtual machine.
Virtual machines that are moved between hosts using the hpvmmigrate command preserve their
UUIDs and are not seen as two virtual machines.
Data collection in a Serviceguard environment
When you collect data from a virtual machine, Capacity Advisor is aware of the guest/host
relationship and collects data from the virtual machine host as well as from the virtual machine.
However, Capacity Advisor does not have knowledge of Serviceguard cluster membership, so
when you collect data when viewing a Serviceguard workload, Capacity Advisor collects data
only from the system you are viewing but not from any other systems in the Serviceguard cluster.
If the workload has executed on any cluster members other than the one you are viewing since
the last time data was collected on all the systems in the cluster, there will be gaps in the data
profile for the Serviceguard workload for those times when the workload was executing on another
cluster member. To fill these gaps, collect data manually from all nodes in the cluster.
158 Using Capacity Advisor with HP Serviceguard