Error while running hostsetup when installing LSF

[rtynd4]

Hi,

I installed Platform_Process_Manager (LSF). When I go to execute this command:

sudo /usr/share/lsfm8/10.1/install/hostsetup --top="/usr/share/lsfm8" --boot="y" --profile="y" --start="y"

I go this error:

Logging installation sequence in /usr/share/lsfm8/log/Install.log

------------------------------------------------------------

    L S F    H O S T S E T U P    U T I L I T Y

------------------------------------------------------------

This script sets up local host (LSF server, client or dynamic host) environment.

[Wed Nov  5 15:38:23 EST 2025:hostsetup:ERROR_6012]

    Host "xxx" is not defined in the cluster.

When setting the host name in the install.config I set these settings:

JS_HOST=xxx.yyy.zzz
LSF_MASTER_LIST="xxx.yyy.zzz"

No where did I set the host name to just the short name xxx.

Please advise how I might fix this issue.

[gwootton]

The script runs the "hostname" command to acquire the local host name and strips the domain if present, so that's where the short name is coming from. The script is checking that hostname against the contents of the lsf.cluster file and not finding a match, causing the error.

You could modify the lsf.cluster file to use the short name to resolve the issue.

[rtynd4]

I see this file: "lsf.cluster.sas_cluster"

Is this the file you are referring to?

I tried adding the host name as just:

Begin Host
HOSTNAME model type server RESOURCES #Keywords
#apple Sparc5S SUNSOL 1 (sparc bsd) #Example
#peach DEC3100 DigitalUNIX 1 (alpha osf1)
#banana HP9K778 HPPA 1 (hp68k hpux)
#mango HP735 HPPA 1 (hpux cs)
#grape SGI4D35 SGI5 1 (irix)
#lemon PC200 LINUX 1 (linux)
#pear IBM350 IBMAIX4 1 (aix cs)
#plum PENT_100 NTX86 1 (nt)
#berry DEC3100 ! 1 (ultrix fs bsd mips dec)
#orange ! SUNSOL 1 (sparc bsd) #Example
#prune ! ! 1 (convex)
xxx.yyy.zzz ! ! 1 (mg)

xxx ! ! 1 (mg)
End Host

But that did not help.

Could I temporarily change the host name to the short name?

Thanks for the help.

[rtynd4]

I did change the lsf.cluster.sas_cluster to use just the short name but got the same error.

[MrSantos]

After editing  lsf.cluster.sas_cluster did you run lsadmin reconfig and badmin reconfig? I believe these are required to recognize the new name defined. 

[rtynd4]

Well :),

I just installed this yesterday and this is my first install.

The processes are not running so of course I get this error:

xxx:/usr/share/lsfm8/conf> badmin reconfig

Checking configuration files ...

There are fatal errors.

Do you want to see detailed messages? [y/n] Do you want to see detailed messages? [y/n] y
Nov 6 11:30:09 2025 124685 3 10.1 getClusterData: ls_getclustername() failed. LIM is down; try later.
Nov 6 11:30:09 2025 124685 3 10.1 minit: getClusterData() failed. LIM is down; try later.
Nov 6 11:30:49 2025 124685 3 10.1 main(): Failed to contact LIM: LIM is down; try later; quit master
---------------------------------------------------------
There are fatal errors.

The server has not been rebooted yet. I may have that done and then try to get the processes running.

Thx

[MrSantos]

If you're installing the first node (or a single one) you probably don't need hostsetup.  This link mentions hostsetup as optional following an install: lsfinstall - IBM Documentation

You can find more information on LSF installation here: Install or upgrade IBM Spectrum LSF - IBM Documentation

[rtynd4]

Some of the files do not have the setuid set on them.

will this command set them?

 sudo /usr/share/lsfm8/10.1/install/hostsetup --top="/usr/share/lsfm8" --setuid

what is top normally set to?

Do I run this for /usr/local/pm too?

[gwootton]

What's in this file?
/usr/share/lsfm8/log/Install.log

[rtynd4]

Not much, just the same output after running the hostsetup command.

[gwootton]

You could add "set -x" to the top of the hostsetup file under "#!/bin/sh", which would write out all the actions the script is performing.

From what I can tell, error 6012 is thrown when the value of $LOCAL_HOST is not in the list of servers and clients it reads from lsf.cluster.cluster-name, if it can find it. If it can't find it, it does check also the LSF configuration directory's hosts file for aliases, if the file exists (e.g. /usr/share/lsfm8/conf/hosts.
So if you changed the lsf.cluster.cluster-name file such that XXX is present there that would make me think that the script is failing to read it at all, or is reading a different file.

[rtynd4]

I was never able to run badmin reconfig because I cannot get lim to start. I run lsfstartup and it runs a long time. I keep killing it. No log files are created. I am going to let it run with the hope it will finally complete and create a lim.log file.

I had to compare the files with an existing installation to determine which files needed the setuid bit set and had the sysadmin manually set them. I did see in the hostsetup script what files to have that set.

I hardcoded the hostsetup LOCAL_HOST to the fqdn but I still get the same error about the host name not defined in the cluster.

Fyi, I am installing this on the same server as an older LSF install exists. I use a different admin account and different paths for the files.

Thx

[gwootton]

The LOCAL_HOST variable is set by the script, setting it beforehand wouldn't change the behavior.

[rtynd4]

I do see when I run lsfstartup this error but it does not exit:

scc8111:/usr/share/lsfm8/conf> sudo -E /usr/share/lsfm8/10.1/sparc-sol10-64/bin/lsfstartup
Starting up all LIMs ...
Do you really want to start up LIM on all hosts ? [y/n]y
Start up LIM on <scc8111.its.state.nc.us> ...... rsh: No such file or directory

[gwootton]

By default, lsfstartup uses the "rsh" command to run commands on other hosts. This error is saying that you don't have rsh installed for that to work.
You can change this command to ssh using the LSF_RSH option in lsf.conf.
https://www.ibm.com/docs/en/spectrum-lsf/10.1.0?topic=lsfconf-lsf-rsh