Administrator's Guide

Failover error handling
Windows Cluster automatically fails over resource groups if the system where resource group is
running on becomes unavailable. This is part of the cluster functionality. Also, this means that if a
problem occurs with the HP 3PAR storage system, a resource group online process will be stopped.
The behavior of HP 3PAR Cluster Extension is highly configurable. Depending upon the customer
setting, Cluster Extension is used to prevent resource groups from going online automatically under
the wrong conditions.
Cluster Extension will return local, data center-wide or even cluster-wide errors to prevent accidental
access to the resource group’s virtual volumes. HP does not recommend restarting a failed resource
group without investigating the problem. A failed Cluster Extension resource indicates the need to
check the status of the Remote Copy volume group and its member virtual volumes and decide
whether it is safe to continue or not.
HP 3PAR Cluster Extension services, resources, or resource groups return a data center error and
fail the resource if the Remote Copy volume group status indicates that the problem experienced
locally would not be solved on another system connected to the same HP 3PAR storage system.
Depending on the resource group and resource property values, the resource tries to start on
different nodes several times. If the remote data center is down, this would look like the resource
group is alternating between the surviving systems. This happens until the previously mentioned
resource and resource group property values are reached or you disable the restarting of the
resource. This could be also the case if the ApplicationStartup resource property has been
set to FASTFAILBACK. If a 3PAR storage system state has been discovered that does not allow
bringing the resource group online on any system in the cluster, a cluster error would be reported
and the resource would fail on all systems. This could lead to the same behavior as described for
a HP 3PAR Cluster Extension data center error.
Failing physical disk resources during online attempt of the resource group
When resource groups that use HP 3PAR Cluster Extension to failover Remote Copy volume group
are brought online, physical disk resources may fail due to the following reasons:
• The physical disk resource does not have a dependency on its HP 3PAR Cluster Extension
resources/packages configured. Review the setup steps for HP 3PAR Cluster Extension
resources.
• The fibre channel path or connectivity between the servers and the storage systems may be
broken. So user has to review the FC connectivity between the servers and the storage systems.
• If the storage array is brought back online or started after the array shutdown due to the
datacenter disaster or Inform OS upgrade, at times the status of the remote copy volume group
may go to the failsafe status as soon as the array is brought back online. The status of the
remote copy volume group is marked as failsafe by the Inform OS after the array comes
online and when the replication roles are primary at one side and primary-rev at the other
side. At this time, the physical disk resource may fail to come online in the Microsoft failover
cluster host whenever the cluster application role tries to come online on the server cluster host
which is connected to the rebooted storage array.
One of the scenarios to get in to the failsafe status can be explained as follows.
The replication roles for a remote copy volume group are primary in one datacenter (primary)
and secondary in the other datacenter (secondary) and the corresponding application in the
Microsoft failover cluster are online in the primary datacenter. In case a disaster such as power
outage happens in the primary datacenter, the application tries to failover to the failover
cluster host in the secondary datacenter. The application comes online successfully in the
failover cluster host in the secondary datacenter if the CLX property UseNonCurrentDataOk
is set to Yes. Once application comes online, the replication role in the secondary datacenter
turns to primary-rev from secondary.
Failover error handling 73