[Condor-users] Very slow file transfers to Grid Sites

[Patrick Armstrong <patricka@xxxxxxx>]

Hi Everyone,

Lately I've been trying to use Condor 7.1.2 to schedule jobs to grid  
sites running Globus 4.0, and I've found that transferring files to  
the grid sites is very very slow. For example, when I run the  
following job, it will stay in the StageIn state for a really long  
time. Anywhere from ten minutes to an hour. I'm only staging a 60 byte  
shell script, and copying the same file with globus-url-copy takes  
less than a second.

Does anyone have any idea what might be going on here?

Here's the job I'm submitting.

Universe = grid
grid_resource = $$(resource_name)
x509userproxy = /tmp/x509up_u510
Transfer_Executable = false
when_to_transfer_output = ON_EXIT_OR_EVICT
globus_resubmit = (My.NumGlobusSubmits =!= UNDEFINED &&  
My.NumGlobusSubmits > 0)  && (My.JobStatus =?= 1)
globus_rematch = true
Executable = test.sh
Requirements = (TARGET.resource_name =!= UNDEFINED)
Transfer_input_files = /home/patricka/test.sh
queue 1

--patrick