Am 28.04.2012 um 06:12 schrieb M Stauffer -V-:

>> Neverthless, if the admin increases the wallclock time, one 
>> could start the tool `screen` and inside the `screen` session 
>> issue a `qlogin`. To this `screen` session  you can detach and 
>> reattach from whereever and whenever you like (supposing it's 
>> issued an always switched on headnode of the cluster).
>> 
>> http://linux.die.net/man/1/screen
>> 
>> Most important: "CTRL-A d" to detach, and later start it with 
>> `screen -xRR` to get the session back.
> 
> I've tried screen a bit before, thanks. Someone else had idea which might 
> work even if the admin doesn't increase wallclock time. To qlogin, *then* 
> start screen and start the debugging process, then detatch and logout. Then 
> qlogin into the *same node* and reattach. I'm going to experiment with that, 
> see if it works.

Well, this would violate the granted scheduling, and AFAICS the screen session 
will be terminated in a proper way due to the attached additonal group ID.

NB: the ownership of the generated /dev/pts/x is wrong and needs to be fixed to 
have access to it as a user (in case you want to test it on your own).

-- Reuti

 

> Cheers,
> Michael 
> 
>> 
>>> Rayson
>>> 
>>> 
>>> On Thu, Apr 26, 2012 at 7:20 PM, "Hung-Sheng Tsao (LaoTsao 
>> 老曹) Ph. D."
>>> <[email protected]> wrote:
>>>> if this is SGE related
>>>> 
>>>> 
>>>> 
>>>> Hi,
>>>> 
>>>> I've been assigned a debugging task on a Rocks 5.4.3 
>> cluster (I helped
>>>> build built the software, but on OSX). I know only the very 
>> basics of
>>>> using the cluster, like making sure I qlogin to run 
>> something instead of
>>>> doing it on the head node, and that I can submit tasks 
>> using qsub, but
>>>> that's about it.
>>>> 
>>>> The app I'm debugging is segfaulting but only on the 
>> cluster, not on my
>>>> Mac, and it will take 20+ hours to segfault judging from the current
>>>> rate. I'm currently running it in gdb via a qlogin session.
>>>> 
>>>> My question is wheter there's a way to start an interactive session,
>>>> then suspend the qlogin session without ending the interactive job
>>>> itself, and then and reattach to the job later on to debug once the
>>>> segfault has happened. The main reason for doing this is that the
>>>> cluster admin has set a 24-hour limit on qlogin sessions. 
>> Also, it'd be
>>>> easier to not have to maintain my terminal connection, but 
>> that's not a
>>>> big issue.
>>>> 
>>>> Or are there other ways to accomplish this? Many thanks for 
>> any help?
>>>> 
>>>> Cheers,
>>>> Michael
>>>> 
>>>> 
>>>> _______________________________________________
>>>> users mailing list
>>>> [email protected]
>>>> https://gridengine.org/mailman/listinfo/users
>>> 
>>> _______________________________________________
>>> users mailing list
>>> [email protected]
>>> https://gridengine.org/mailman/listinfo/users
>>> 
>> 
> 
> 


_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users

Reply via email to