Configuring PATROL Agent in a UNIX cluster

The information in this section provides a general idea of the processes involved in setting up a UNIX cluster environment and integrating PATROL into that environment. Procedures and steps describing how to set up third-party software are intended as a general outline of the process for that product and are not intended as step-by-step instructions.

Setting up PATROL to run in a UNIX cluster environment consists of several standard tasks. The standard cluster administration tasks and the PATROL-specific tasks are described in general terms. This section provides a high-level overview of building a UNIX cluster and integrating PATROL into that environment.

This topic details the following information:

Installing the application on a shared disk

Install the cluster application on a shared disk that's available to the entire cluster. In UNIX environments, the executable can be installed on the local disk. Such an installation strategy may require additional administrative and maintenance tasks.

Installing the PATROL Agent on each cluster node

Use the same installation strategy for the PATROL Agent that you used for the cluster application.

Note

You must have at least one agent executable installed on the node's local drive to monitor the node's operating system. If you select to install the cluster application on the local disk, you'll also have one to monitor the cluster application.

Assigning a unique port number

During agent installation on each node, assign a unique, listening port number to the PATROL Agent bound to the cluster application. If you install the agent on each node, the port must be the same across all nodes within the cluster.

Defining the PATROL cluster-specific environment variables

In this task, you define the PATROL cluster-specific environment variables in the .profile (Korn and Bourne shell) or .cshrc file (for C shell) of the account used to start the agent. This action ensures that the agent reads configuration information and writes history information to the same set of files, regardless of which host it is running on.

Perform the following task on each node in the cluster.

For Korn and Bourne shells, use the following format:

PATROL_VIRTUALNAME_<port>=<virtual server name>
PATROL_HISTORY_<port>=</History_Directory>
PATROL_CONFIG_<port>=</Config_Directory>

export PATROL_VIRTUALNAME_<port>
export PATROL_HISTORY_<port>
export PATROL_CONFIG_<port>

For C shell, use the following format:

setenv PATROL_VIRTUALNAME_<port> <virtual server name>
setenv PATROL_HISTORY_<port> </History_Directory>
setenv PATROL_CONFIG_<port> </Config_Directory>

For more information about specific variables, see PATROL cluster-specific environment variables for history and configuration.

Defining the PATROL Agent as a member of the package

This section explains how to add the PATROL Agent service to the cluster application's control script for a package. This task is commonly referred to as binding the agent to the cluster application.

Note

The following code is a subset of the control script and does not represent a complete script. It is provided to give you an idea of the type and extent of changes you must make in HP's MC ServiceGuard or other UNIX cluster management software.

Make the following additions to the customer_defined_run_cmds:

# Start of user-defined functions.
function customer_defined_run_cmds
{
# Add customer defined run commands.
# PATROL Start command (substitute correct path
# and port).
\\ \\ su patrol -c "/opt/PATROL3.4/PatrolAgent -p 3939"
# End of Patrol Start command.
test_return 51}}
}

Make the following additions to the customer_defined_halt_cmds:

function customer_defined_run_cmds
{
# Add customer defined halt commands.
# PATROL halt command (substitute correct path
# and port).
PID=$ (ps -ef|awk '$NF=="3939" \{ print $2\}')
if\[[ -n $PID ]]
then
kill -TERM $PID
fi
# End of Patrol Start command.
test_return 52
}