HSG80 ACS Solution Software V8.6 for IBM AIM Installation and Configuration Guide

Installing and Configuring the HS-Series Agent 4–27
Node Isolation Recovery
When any single node experiences a loss of services in a multi-node cluster such that it
loses all contact with the rest of the cluster yet still remains up with a running Agent, it is
considered an “isolated” node. It is important that the Agent process be killed prior to
starting another instance of the Agent on another node. This scenario is handled by
appending code to the node_down_local event script, to shut down the Agent
unconditionally. During Agent installation, this code is automatically appended to the
node_down_local event script.
Failure of the Agent
Failure of the Agent during an HACMP session is handled by the HACMP event scripts.
This is achieved by placing the Client IP address into a Resource Group and starting the
server_down event script upon failure of the Agent process.
Polling all Nodes in Cluster for Presence of a Running
Agent
When the Agent fails, HACMP invokes the start_server event script. This script will
restart the Agent automatically using the HACMP_Config.sh script, Option 25.
Restarting the Agent on Surviving Node
When the HACMP software completes the process of failover after a NODE_DOWN event.
The start_server event script will restart the Agent automatically using the
HACMP_Config.sh script, Option 25.
Notification
Notification type and level are determined by the settings of these parameters in the
configuration of the surviving node at the time the Agent process is restarted. These
parameters may or may not be the same for every node in the cluster.
Supporting Tools
The HACMP_install.sh installation script is called directly by the install.sh
script and performs all necessary event modifications to the HACMP event scripts on the
local node. This process must be repeated for each node in the cluster.