The usage of LSF hosts file

Answer

INTRODUCTION:
LSF identifies a host as part of LSF cluster (client and server) by its name so LSF needs to match host alias or address with an unique host name. LSF looks up a host name in the following order.  The resolved official host names will be displayed and used by LSF.

    In the LSF_ENVDIR/hosts file

    In the /etc/hosts file

    Sun Network Information Service/Yellow Pages (NIS or YP)

    Internet Domain Name Service (DNS).

Both forward lookup (host name -> IP) and reverse lookup (IP-> host name) should be supported by the environment on LSF hosts.


SCOPE:
LSF Hosts file is a change that only affects LSF cluster to workaround some environment issues and does not most of the times require any  IT  change request .

PURPOSE:
LSF_CONFDIR/hosts file makes LSF bypass the configurations/settings on system level. In the following cases, you may need to use LSF hosts file

- For special name convention for host names within LSF, LSF_CONFDIR/hosts file is necessary.

- A multi-homed host which has multiple IP's with different host names and each interface can then be directly accessed by name. In this case, one official host name must be configured so that all of the Internet addresses for a host resolve to the same host name. If customer doesn't want to modify the system hosts file (/etc/hosts) to affect the whole system, LSF hosts file is the way to make LSF the only application that resolves the addresses to the same host.

- When reverse DNS lookup is not supported for client hosts

STEPS:
The format of LSF_CONFDIR/hosts is the same as for /etc/hosts as the following.
<IP address>        <official host name>         <alias 1> <alias 2> ... <alias n>

Official host name: The first name configured for each address is called the official name.
alias:  Other names for the same host are called aliases.

Example 1 for single IP address:
AA.AA.AA.AA  host-AA host-1AA host-2AA

Example 2 for multiple IP addresses:
AA.AA.AA.AA  host-AA-BB  host-1AA host-2AA ... # first interface
BB.BB.BB.BB  host-AA-BB  host-1BB host2BB ...  # second interface

lim, res, sbatchd and mbatchd need to be restarted to pick up the contents in LSF $LSF_ENVDIR/hosts file for host name resolution.


EXTRA RESOURCES:
·        Troubleshooting document for LSF Network related issues:http://www-01.ibm.com/support/docview.wss?uid=isg3T1023608
·        "Administrating IBM Platform LSF" guide for more details about hosts with multiple addresses in session "Hosts with multiple addresses"