If you have an EGO cluster with LSF running, you can install as an EGO service. This transfers control of Process Manager to EGO through the Platform Management Console. The primary advantage is the ability to fail over Process Manager server to another host.
For a default installation on either Windows or UNIX, follow the applicable installation instructions in the following guides:
If you have chosen the default installation in a Windows environment, you may want to set up failover.
If you are installing with the default installation option and want to enable Windows failover, you must enable JFD failover on Windows.
Decide which host will be the primary host, and which host will be the fail-over host. For example, the primary host is HOSTP, the failover host is HOSTF.
Install LSF cluster. Both HOSTP and HOSTF should belong to the cluster. If you already have an LSF cluster installed, add HOSTP and HOSTF to the cluster.
Make sure LSF administrator is in Administrators group on both primary host (HOSTP) and failover host (HOSTF), and has all required privileges:
Make sure the LSF administrator has read/write permission on %JS_HOME% and %LSF_TOP%\conf\passwd.lsfuser.
Start to install Process Manager on the primary host, but make sure to specify a shared directory as the Destination Folder. The path should use a UNC (Universal Naming Convention), not a mapped drive. For example, %JS_HOME% could be \\HOSTS\ppm7.
By default, Process Manager is installed as a Windows service. After the installation is complete, change the startup type of Process Manager from Automatic to Manual to prevent it from being started up automatically when the machine reboots. From the Platform Management Console, stop the Process Manager service first if it is started.
Modify/Add system environment variables on both HOSTP and HOSTF.
On the primary host, all environment variables were already set during installation.
Configure a queue dedicated for jfd failover in lsb.queue.
RERUNNABLE = yes: JFD runs as a rerunnable job. If it is not set to yes, failover will not work.
PREEMPTION = PREEMPTIVE: Makes jfd_failover queue a preemptive queue. Even if JFD is submitted to LSF as a regular job, it should have high priority. When JFD and other jobs compete for the same resource (for example, a job slot), it should be able to preempt low-priority jobs.
REQUEUE_EXIT_VALUES = 1 2 3 4 5 6 7 9 127 128 255: JFD is automatically requeued when it exits with the specified exit values:
JOB_CONTROLS = TERMINATE [jadmin stop]: When JFD is being terminated with bkill, the actual action will be jadmin stop so the JFD can exit cleanly.