Symantec logo

Automating local cluster failover for a bunker

This section describes how to set up the VCS agents to automate failover of the bunker when the Primary fails over within a local cluster. This step is not required if the bunker is set up using the IP protocol. For IP protocol, the bunker setup is the same whether the Primary is a single node or a VCS cluster.

When a bunker is set up using the STORAGE protocol, the disk group containing the bunker RVG is imported on the Primary host. If the Primary RVG is in a VCS cluster, the bunker RVG must remain online on the same node on which the parent application RVG is online.

In a private disk group environment, the RVG resource handles the failover process. If the host on which the RVG resource is online fails, the RVG resource fails over to another host within the cluster. The RVG resource ensures that the bunker RVG also fails over, so that the bunker RVG continues to be on the same host with the parent RVG.

In a shared disk group environment, the application RVG and the bunker RVG must be online on the logowner host. If the logowner fails over, the bunker RVG must be deported from the original logowner host and imported on the new logowner host. In this case, the RVGLogowner agent handles the failover process.

To set up automated failover of the bunker RVG, specify the bunker RVG, the bunker disk group, and the bunker host using the following attributes of the RVG resource in the application service group or the RVGLogowner agent:

Attribute

Description

StorageDG 

The name of the bunker disk group.  

StorageRVG  

The name of the bunker RVG.  

StorageHostIds 

A space-separated list of the hostids of each node in the bunker cluster.  

The above attributes are the only specific attributes that differ for an RDS containing a bunker. The rest of the configuration for the VCS Agents for VVR is the same as for any other RDS. See the Veritas Cluster Server Agents for Veritas Volume Replicator Configuration Guide for more information.

Using the StorageHostIds attribute

If the bunker site is a cluster, make sure that the bunker RVG group is never online when the bunker disk group is imported on the Primary cluster. Otherwise, the bunker disk group would be imported on two hosts at the same time, which results in split brain.

If an automatic failover occurs in the Primary cluster, the agent refers to the StorageHostIds attribute to help ensure that the bunker RVG is not imported on both a bunker host and a host in the Primary cluster at the same time. The Primary cluster does not import the bunker disk group if the disk group is already imported on a host in the bunker cluster.

To determine the hostid, issue the following command on each node:

# vxdctl list

Volboot file

version: 3/1

seqno: 0.5

cluster protocol version: 60

hostid: vvrnode1

defaultdg: pdg

If the hostid of a bunker node changes, you must change the StorageHostIds attribute to reflect the new value using the following command:

# hares modify RvgPriResName StorageHostIds value