[GE users] Strange behaviour of environment variables

Andy Schwierskott andy.schwierskott at sun.com
Wed Sep 29 15:04:08 BST 2004


Terry,

that's a good insight!

Indeed - looking at the code in the shepherd it seems that SGE adds the
environment variables to the existing environment instead of setting up a
new environment.

Andy

> OK,  I been receiving some interesting emails from Olle Liljenzin
> and he is confirming my suspicions.  Please see the following
> log which seems to confirm that some environment variables are
> inherited from the process that starts execd:
>
> {117}{tlalonde at wstlalonde}> qrsh -l hostname=ren echo \$XYZZY
> XYZZY: Undefined variable.
> {118}{tlalonde at wstlalonde}> rsh ren
> {102}{tlalonde at ren}> su -
> Password:
> root has logged on pts/0 from XXX.XXX.XXX.XXX:0.0.
> [root at ren]# /etc/init.d/sgeexecd stop
>   Shutting down Grid Engine execution daemon
> [root at ren]# setenv XYZZY blah
> [root at ren]# /etc/init.d/sgeexecd start
>   starting sge_execd
> [root at ren]# exit
> {103}{tlalonde at ren}> exit
> logout
> Connection closed.
> {119}{tlalonde at wstlalonde}> qrsh -l hostname=ren echo \$XYZZY
> blah
>
>
> -----Original Message-----
> From: Terry Lalonde
> Sent: Tuesday, September 28, 2004 9:46 AM
> To: users at gridengine.sunsource.net
> Subject: RE: [GE users] Strange behaviour of environment variables
>
>
> Man, I searched high a low and cannot find out where
> that env-variable is being set. And, Why did restarting the
> execd on one host change the behaviour on that host?
> It remains a mystery to me.
>
> -----Original Message-----
> From: Andy Schwierskott [mailto:andy.schwierskott at sun.com]
> Sent: Tuesday, September 28, 2004 4:46 AM
> To: users at gridengine.sunsource.net
> Subject: RE: [GE users] Strange behaviour of environment variables
>
>
> Terry,
>
>> The HSPICE environment variable is not in the "environment" file.
>> So in that case where does/did the value come from?  the execd
>> process?
>
> No, definitely not. It must be well hidden in one of your shell profile
> files.
>
> Check /etc/*sh* /etc/.*sh* and all files which are sourced by these files
> (samep of course applies for the dot files in the home directory.
>
> Andy
>
>> Regardless, using the -V option results in the desired behaviour. I
>> added it to our sge_request file, because I cannot think of a case where
>> I would not want that behaviour (at least in our environment).
>>
>> Thanks,
>> Terry
>>
>> -----Original Message-----
>> From: Andy Schwierskott [mailto:andy.schwierskott at sun.com]
>> Sent: Monday, September 27, 2004 1:21 PM
>> To: users at gridengine.sunsource.net
>> Subject: RE: [GE users] Strange behaviour of environment variables
>>
>>
>> Terry,
>>
>> could you should check the "environment" file in the shepherd directory of
>> the job. If HSPICE/SPICE is not defined there it's not related to the
>> submission machine, otherwise the setting could be done in a ".sge_request"
>> file.
>>
>> "qstat -j <jobid>" will show all variables exported with "-v" or "-V"
>>
>> Andy
>>
>>
>>> Yes sorry, that was a typo  the SPICE strings should had read HSPICE.
>>>
>>> Terry
>>>
>>> -----Original Message-----
>>> From: Reuti [mailto:reuti at staff.uni-marburg.de]
>>> Sent: Friday, September 24, 2004 3:31 PM
>>> To: users at gridengine.sunsource.net
>>> Subject: RE: [GE users] Strange behaviour of environment variables
>>>
>>>
>>>> I just did this and it yielded nothing.
>>>
>>> And the appearance of $HSPICE and $SPICE is a typo? Another place to define it
>>> could be a "-v HSPICE=2004.03-SP1" line in a default sge_request. - Reuti

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