HP 3PAR Geocluster Software - Manual (Page 13)

3PAR GeoCluster 1.0.0.14 for Windows User’s Guide
320-200263 Page 13
CAUTION: See section 3.4 Best Practices for considerations related to the
use of SSH.
2.2 3PAR GeoCluster for Windows Failover Algorithm
This section describes the method 3PAR GeoCluster for Windows follows in order to handle
failovers.
2.2.1 Failover Sequence from a Microsoft Cluster Server Perspective
Failover occurs from one of the following:

An administrator's action results in an active node becoming unavailable.

One of the nodes’ cluster resources goes offline.
You can set whether a cluster resource failure will trigger a failover in its properties.
When a failover is triggered, it causes MSCS to bring the whole resource group offline and
search for a functional node to bring the resource group online. If the dependencies are set
correctly, MSCS will attempt to bring the 3PAR GeoCluster for Windows resource online.
The 3PAR GeoCluster for Windows resource will perform a decision algorithm, taking into
account conditions such as whether it is on the primary or secondary site and whether the
Remote Copy links are up.
Depending on the outcome of the decision process, that particular node may be able to
bring its 3PAR GeoCluster for Windows resource online. If online, the cluster will move to
the next node and so on, until it is able to find a node able to start its 3PAR GeoCluster for
Windows resource and the resources that depend on it or until it fails.
2.2.2 Failover Sequence from a Remote Copy Perspective
In a normal working situation, the InServ storage arrays and the Remote Copy groups they
host are considered by the 3PAR GeoCluster for Windows resource to have one of the
following roles:

Primary - This array is considered to be the replication’s source, when the cluster is in
normal working order.

Secondary - This array is considered to be the replication’s destination, when the
cluster is in normal working order.
If the Remote Copy replication fails, whether due to a failure or administrator action, and
MSCS has a node try to take hold of the replication VVs on the secondary InServ , then the