[GE users] Bug in sge_schedd/sge_qmaster

Reuti reuti at staff.uni-marburg.de
Wed Dec 7 16:28:50 GMT 2005


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi,

Am 07.12.2005 um 16:21 schrieb Sebastian:

> Hello,
>
> i think i found a bug, so please help me.
>
> First a short descprition:
>
> I run the sge_schedd and sge_qmaster on a lx24_x64-machine, the  
> sge_execd
> are running on lx24_amd64-machines.
>
> I integrated everey thing as written in Reuti's HowTo for LAM- 
> Integration.
>
> I made a local .bashrc for root/the sge-admin and the submittion- 
> user on the
> qmaster-node and the two execd-nodes, where i set the Source-Script  
> for SGE
> and LAM/MPI.
>
> When i submit a simple job as user where the LAM-Enviroment should be
> started and the command "lamnodes" should be executed, a error was  
> reported,
> because the Environment-Variables aren't take over.
>
> I implemented an echo-command for the $PATH-Variable in the  
> startlam.sh and
> there the wrong entries were reported, not the entries which i made  
> in the
> .bashrc.

the .bashrc isn't sourced there. Only for a login (i.e. via qrsh)  
it's sourced. If you need some variables for any reason in  
startlam.sh, you could source it therein by hand:

. ~/.bashrc

in the script before the universe is built. If you need this for any  
license-server AFAIR, then it might be easier to replace the original  
program call with a script setting these fixed variables. So maybe  
you don't need to touch .bashrc or need -V at all. E.g. in pseudo-code

#!/bin/sh
export LICENSE_SERVER=srv
export LICENSE_MODE=full
exec the_program.orig

and name this script the_program. It won't hurt normal operation, and  
you don't have to worry about a set LICENSE_SERVER in SGE at all.

> Now the question is, in which Environment is the startlam.sh-Script
> submitted, it is told the PE as start_proc_args.
>
> This step has a little bug or i'm the bug, because there the  
> environment
> isn't set as defined.
>
> When i made a entry "#$ -V" in the job-script, all is made correct,  
> but the
> needed LAM-Structure isn't made, on the Slave-Nodes the LAM-Process  
> is no
> child of sge_execd eg. sge_sheperd.

Are you extending the $PATH there or replacing it completely? The  
first entry should be the $TMPDIR, so that the rsh-wrapper is found  
first before any other rsh in the system.

Cheers - Reuti

> Now how can i make this right or isn't there any possibility?
>
> Thanks and Kind Regards,
>
> Sebastian
>
> -- 
> 10 GB Mailbox, 100 FreeSMS/Monat http://www.gmx.net/de/go/topmail
> +++ GMX - die erste Adresse für Mail, Message, More +++
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
> For additional commands, e-mail: users-help at gridengine.sunsource.net


---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
For additional commands, e-mail: users-help at gridengine.sunsource.net




More information about the gridengine-users mailing list