[GE users] SGE on Debian-64 interactive qrsh without limits

Reuti reuti at staff.uni-marburg.de
Mon Jan 16 20:02:33 GMT 2006


Am 15.01.2006 um 16:37 schrieb Jan Behrend:

> Hi Reuti,
>
> Reuti wrote:
>> Hi, Jan,
>>
>> thanks for your reply. We are not using any starter_method,
>
> I think you got me wrong. I was trying to point out that the  
> environment
> problems (missing or not complete $PATH etc.) could relate to the
> missing "login -- <user>" process on Debian systems, and as a  
> result of
> this, one had to use the starter_method (the "#!/bin/bash -l ...  
> exec $*
> ..." wrapper stuff).  But maybe I am mistaken ...
>

The $PATH and other settings are ok. Just the limits are wrong. But I  
found something:

from pam_0.76-22.diff of Debian:

--- pam-0.76.orig/debian/patches-applied/ 
027_pam_limits_better_init_allow_explicit_root
+++ pam-0.76/debian/patches-applied/ 
027_pam_limits_better_init_allow_explicit_root
@@ -0,0 +1,100 @@
+Allow explicit limits for root.
+Also, remove limits on su.
+Index: Linux-PAM/modules/pam_limits/pam_limits.c

Seems to be a Debian-feature. I'll recompile pam_limits.so after  
changing it back and post the result here. - Reuti


> Cheers Jan
>
>> and I get
>> the impression that this is a Debian problem as it resets the limits
>> and not SGE's fault. I also found, that every user on a Debian-64
>> system can reset the limits by doing a dchroot to the 32-bit   
>> environment:
>>
>> $ ulimit -t 55
>> $ ulimit -aH
>> ...
>> cpu time             (seconds, -t) 55
>> ...
>> $ dchroot
>> Executing shell in 'ia32' chroot.
>> $ ulimit -aH
>> ...
>> cpu time             (seconds, -t) unlimited
>> ...
>>
>> in my opinion this shouldn't be. Here dchroot seems to be the  
>> source  of
>> the problem, and for qrsh maybe a patched in.rlogin.d. I'll turn   
>> over
>> to the Debian list and post the results here.
>>
>> Thx again - Reuti
>>
>>
>> Am 15.01.2006 um 13:55 schrieb Jan Behrend:
>>
>>> Reuti wrote:
>>>
>>>> Hi all,
>>>>
>>>> with our new Opteron cluster we try to use Debian. Before we  
>>>> used   SuSE
>>>> on the 32 bit machines. One issue I have are the missing hard-  and
>>>> soft-limits for interactive jobs. With batch jobs the limits I  can
>>>> check successfully with ulimit in the script. So I wonder,    
>>>> whether
>>>> this
>>>> is a Debian or SGE (configuration-)problem, and would  like to  
>>>> get  from
>>>> SuSE-64 and Debian-32 users the statement, which  behavior they   
>>>> see, as
>>>> I have only SuSE-32 and Debian-64 accessible.  (SuSE 9.3, SGE  
>>>> 6.0u7,
>>>> Debian 3.1)
>>>>
>>>> ** On  SuSE 32-bit.
>>>>
>>>> $ qrsh -l h_cpu=55
>>>> Last login: Thu Jan 12 13:21:36 from login
>>>> Have a lot of fun...
>>>> reuti at node02:~> ulimit -aH
>>>> ...
>>>> cpu time             (seconds, -t) 55
>>>> ...
>>>>
>>>> 26874 ?        S      0:00  \_ sge_shepherd-7339 -bg
>>>> 26875 ?        Ss     0:00      \_ /usr/sbin/in.rlogind
>>>> 26876 ?        Ss     0:00          \_ login -- reuti
>>>> 26877 pts/0    Ss     0:00              \_ -bash
>>>> 27010 pts/0    R+     0:00                  \_ ps -e f
>>>>
>>>>
>>>> ** on Debian 64-bit
>>>>
>>>> $ qrsh -l h_cpu=55
>>>> Last login: Fri Jan 13 15:56:08 2006 from node0.cluster on pts/0
>>>> Linux node013 2.6.15 #1 SMP Thu Jan 5 14:04:21 CET 2006 x86_64 GNU/
>>>> Linux
>>>> reuti at node013:~$ ulimit -aH
>>>> ...
>>>> cpu time             (seconds, -t) unlimited
>>>> ...
>>>>
>>>> 2936 ?        S      0:00  \_ sge_shepherd-231 -bg
>>>> 2937 ?        Ss     0:00      \_ /usr/sbin/in.rlogind
>>>> 2938 pts/0    Ss     0:00          \_ -bash
>>>> 3001 pts/0    R+     0:00              \_ ps -e f
>>>>
>>>> Note the missing "login -- reuti" on Debian.
>>>>
>>>> (The command line "qrsh -l h_cpu=60 ulimit -aH" is working   
>>>> correctly.)
>>>>
>>>> Thx - Reuti
>>>>
>>>> ------------------------------------------------------------------- 
>>>> --
>>>> To unsubscribe, e-mail: users-unsubscribe at gridengine.sunsource.net
>>>> For additional commands, e-mail: users- 
>>>> help at gridengine.sunsource.net
>>>>
>>> Hi Reuti,
>>>
>>> this is what I get on Debian 32-bit:
>>>
>>> $ qrsh -verbose -l h_cpu=55 -l arch=lx24-x86
>>> Last login: Sun Jan 15 13:36:26 2006 from
>>> beowulf-vpn.beowulf.iri.mpifr-bonn.mpg.de on pts/0
>>> Linux iripc80 2.6.8-2-686-smp #1 SMP Tue Aug 16 12:08:30 UTC 2005  
>>> i686
>>> GNU/Linux
>>>
>>> $ ulimit -aH
>>> ...
>>> cpu time             (seconds, -t) unlimited
>>> ...
>>>
>>> $ ps -e f
>>> 12189 ?        S      0:00  \_ sge_shepherd-1154 -bg
>>> 12190 ?        Ss     0:00      \_ /usr/sbin/in.rlogind
>>> 12191 pts/0    Ss     0:00          \_ -bash
>>> 12253 pts/0    R+     0:00              \_ ps -e f
>>>
>>> By the way, is this relatet to the sge starter_method (Debian?)   
>>> problem:
>>> http://gridengine.sunsource.net/servlets/ReadMsg?
>>> listName=users&msgNo=13733
>>> or
>>> http://gridengine.sunsource.net/servlets/ReadMsg?
>>> listName=users&msgNo=8503
>>>
>>> Cheers Jan
>>>
>>> -------------------------------------------------------------------- 
>>> -
>>> 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
>>
>
> ---------------------------------------------------------------------
> 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