The Installation Guide no longer applies directly to the OSS release because the INSTALL script is no longer needed, but you still may find it useful. In general, you will want to install the RPMs as follows:
pbspro-server: This should be installed on one node (or two if you employ failover) in your cluster where the server and scheduler will run. Most admins prefer not to run jobs on this node, but it is certainly possible. This RPM also installs everything you would get with the execution and client packages.
pbspro-execution: This package should be installed on all nodes where you want your jobs to run. It provides the pbs_mom daemon, but not the server or scheduler. Admins generally restrict login access to these nodes to prevent users from logging in and impacting running jobs. This RPM also includes everything you would get with the client package.
pbspro-client: This package should be installed on nodes where users are allowed login access, like their workstations. It does not contain any daemons, but allows users to submit and control their jobs.
pbspro-debuginfo: Only required when trying to debug an issue via gdb. You won’t need this unless you manage to crash something in PBS Pro that creates a core dump.
pbspro-*-src: This is the source code for PBS Pro, also used for debugging.
Good luck, and let us know if we can provide additional advice!
Thanks for your reply. First of all, can i know any specific guide for setup PBS Pro in linux cluster environment? Any specific directory needed to be NFS share among the cluster as i used to be using LSF that require to use /home directory to be share.
I did install the RPM following your guide at server and execution host accordingly. Any specific command i can ensure these nodes are properly setup as cluster?
PBS Pro does not require a shared filesystem to function unless you are employing the server failover feature. Chapter 1 of the installation guide provides some illustrations of possible configurations. Step 12 in the quick start guide provides some commands you may use to test your PBS Pro configuration. Running “pbsnodes -a” will show the status of your execution nodes. The state of each node should be “free” when there are no jobs running.
If user IDs are common across your cluster (perhaps you use NIS, LDAP, etc.) you will want to enable “flatuid” on the server. If you utilize a shared filesystem on your cluster, you will want to add “$usecp” to the mom_priv/config files. Details of these and many other features are provided in the admin guide.
In case anyone is wondering, here is the simple way to install pbspro-client… for version 14, and most notably, the packages needed before that. The pbspro-client is needed on nodes which are neither PBS server nor PBS execution nodes. The pbspro-client install is needed on nodes which needs the pbs commands, eg qsub, qstat, that allows users to submit and monitor jobs.
Install PBS client (commands)
[root@…]# rpm -ivh --test pbspro-client-14.0.1-0.x86_64.rpm
error: Failed dependencies:
libXft.so.2()(64bit) is needed by pbspro-client-14.0.1-0.x86_64 libtcl8.5.so()(64bit) is needed by pbspro-client-14.0.1-0.x86_64 libtk8.5.so()(64bit) is needed by pbspro-client-14.0.1-0.x86_64
perl(Env) is needed by pbspro-client-14.0.1-0.x86_64
perl(Switch) is needed by pbspro-client-14.0.1-0.x86_64
*** Postinstall script called as follows:
*** /opt/pbs/libexec/pbs_postinstall client 14.0.1 /opt/pbs
*** No configuration file found.
*** Creating new configuration file: /etc/pbs.conf
*** =======
*** NOTICE:
*** =======
*** The value of PBS_SERVER in /etc/pbs.conf is invalid.
*** PBS_SERVER should be set to the PBS Pro server hostname.
*** Update this value before starting PBS Pro.
*** Replacing /etc/pbs.conf with /etc/pbs.conf.14.0.1
*** /etc/pbs.conf has been created.
*** The PBS commands have been installed in /opt/pbs/bin.
*** End of /opt/pbs/libexec/pbs_postinstall
As root, edit the file /etc/pbs.conf, following the instructions above.