4.1.0 HP PolyServe Matrix Server administration guide (T5392-96073, October 2010)

4. If the virtual host is currently active on a server and that virtual host has the
NOFAILBACK policy, then this active server is moved to the head of the list of
preferred servers to be considered.
5. From this list, ClusterPulse tries to find a server with all services up and enabled.
If ClusterPulse finds a server meeting these conditions, it will use it, preferring
servers earlier in the list of servers configured for the virtual host.
6. If there are no servers with completely healthy services, ClusterPulse picks a
server that has at least one service up and enabled. If the services have different
priorities, ClusterPulse uses a server where a service with the highest available
priority is up. If ClusterPulse finds a server meeting these conditions, it will use
it, preferring servers earlier in the list of servers configured for the virtual host.
7. If there are no servers with an up and enabled service and the virtual host is
configured with the “Always active” option, ClusterPulse picks any server from
the list.
8. The selected server will have one interface that was configured for this virtual
host. The virtual host will be active on this interface. If ClusterPulse cannot locate
a server meeting these conditions, it does not place the virtual host anywhere.
Customize service and device monitors for failover
By default, when a service or device monitor probe fails, indicating that the watched
service is down or the monitored device cannot be accessed, ClusterPulse will fail
over the associated virtual host to another server where the monitored service or
device is up.
You can customize this behavior using the Advanced monitor settings.
Specify failover/failback behavior
The Probe Severity setting allows you to specify whether a failure of the service or
device monitor probe should cause the virtual host to fail over. For example, you
could configure a custom device monitor to watch a router. The device monitor probe
might occasionally time out because of heavy network traffic to the router; however
the router is still functioning. In this case, you can configure the device monitor so
that the virtual host will not fail over if only the device monitor has failed.
The Probe Severity setting also affects the failback behavior of the virtual host.
There are three settings for Probe Severity:
• NOFAILOVER. The virtual host does not fail over when a monitor probe fails.
Configure virtual hosts198