[torqueusers] Yet another problem of stage-in

Guillaume ALLEON guillaume.alleon at laposte.net
Tue Oct 18 14:21:21 MDT 2005


>Are you talking about the "stagein" and "stageout" files (configured
>with qsub's -W arg), or delivering the stdout/stderr of the job
>(configured with qsub's -o, -e, -k, and -j options)?
>  
>
I am talking about both of them.

>The logic for delivery of stdout/stderr files is as follows:
>
>First, if the file(s) are to be delivered to a local filesystem, then
>'cp' is used.  If the file(s) are to be delivered to a remote filesystem
>covered by MOM's $usecp option, then use 'cp'.
>  
>
So a mom_priv/config with usecp means that for the specified filesystem 
cp must e used instead of either pbs_rcp or scp.

>Otherwise, if ./configure was run with --with-scp, then MOM attempts to
>use 'scp', else use 'rcp'.
>  
>
That's my setting.

>When using 'scp', MOM retries with 'mom_rcp' if the scp fails.
>
>Note that CVS just changed last week, it will no longer fallback to
>mom_rcp.
>
>  
>
Doing some ps on the node shows me a pbs_rcp hanging because rcp is not 
allowed on my cluster. The problem is that in the log I got no trace of 
the scp (I was getting such trace in torque 1.1) so I can't figure out 
why scp is not working.

Trying to perform a scp manually as a "simple" user is actually working 
fine.

I "just" would like to have a methodology to trace my problem since I 
know that the cause for such a problem can be very stupid  (and I am 
sure it is  since my other cluster is working like a charm).

Thanks
Guillaume

> 
>  
>
>>2-Is there a Torque/Maui ... BOF/meeting at SC this year ;-)
>>    
>>
>
>If not officially, we'll do _something_.
>  
>
Cool !

>  
>
>------------------------------------------------------------------------
>
>_______________________________________________
>torqueusers mailing list
>torqueusers at supercluster.org
>http://www.supercluster.org/mailman/listinfo/torqueusers
>  
>



More information about the torqueusers mailing list