Re: [Condor-users] File stage-in fails in Condor 6.9.1 as a submitter to GT4

[Jaime Frey <jfrey@xxxxxxxxxxx>]

On Jan 23, 2007, at 6:18 PM, Gabriel Mateescu wrote:

> There is a problem with file stage-in that appears in Condor-G
> for Condor version 6.9.1, but did not appear in Condor 6.7.x.
>
> Specifically, the following submit file
>
> $ more  host.condorg
> universe  = grid
> grid_resource = gt4 FQDN:8443 Fork
> Executable = /bin/hostname
> InitialDir = /home/gabriel/SubmitCondor
> Output =   host.condorg.$(Cluster).out
> Error  =   host.condorg.$(Cluster).err
> Log    =   host.condorg.$(Cluster).log
> log_xml = True
> Notification = Never
> Transfer_Executable = False
> when_to_transfer_output = ON_EXIT_OR_EVICT
> queue
>
>
> works in Condor 6.7.x with FQDN being the same host as the
> central manager (i.e., Condor and Globus GRAM are co-located),
> as well as being a remote host (i.e., submit to a
> remote Globus resource).
>
> However, in Condor 6.9.1, it only works when FQDN is
> the local host, i.e., Condor submits to the Globus GRAM
> on the same machine as the Condor central manager.
> If FQDN is a a remote machine, the job is put on hold and
> the following error occurs
>
>  HoldReason = "Globus error: Staging error for RSL element  
> fileStageIn."
>
>
> In both cases, in the job class-ad I see
>
>   x509userproxy = "/tmp/x509up_u501"
>
> where
>
>   $ grid-proxy-info  -type -timeleft
>     Proxy draft (pre-RFC) compliant impersonation proxy
>     571631
>
>
> I think that before submitting the job, Condor-G delegates
> the X509 credential, then inserts the EPR of the delegated
> credential resource in the job RSL submitted to Globus.
> Has anything in the credential handling changed between
> versions 6.7.x and 6.9.1 of Condor?

Using your submit file, I can successfully submit a job to a remote  
machine using Condor 6.9.1. Nothing has changed in how Condor  
delegates the job credential between 6.7.x and 6.9.1.

If you set 'GRIDMANAGER_DEBUG = D_FULLDEBUG' in your config file, you  
should see a large java exception stack trace in the gridmanager log  
for the stage-in error. It should have more information about what  
exactly went wrong.

+--------------------------------+-----------------------------------+
|           Jaime Frey           | I used to be a heavy gambler.     |
|       jfrey@xxxxxxxxxxx        | But now I just make mental bets.  |
| http://www.cs.wisc.edu/~jfrey/ | That's how I lost my mind.        |
+--------------------------------+-----------------------------------+