[torqueusers] Slow response of torque when jobs are running

Ken Nielson knielson at adaptivecomputing.com
Tue Dec 8 09:59:18 MST 2009


Luc,

In addition to the comments already sent by Josh and Garrick, I just 
wanted to pass this along. I took the following from the trace.

write(3, "+2+12+19+3luc+0+0+0", 19)     = 19 <0.000004>
poll([{fd=3, events=POLLIN|POLLHUP, revents=POLLIN}], 1, 10800000) = 1 
<86.064923>
fcntl(3, F_GETFL)                       = 0x2 (flags O_RDWR) <0.000003>
read(3, "+2+1+0+0+62+55+22+1443402.gweyri"..., 262144) = 85573 <0.000049>

The write is requesting PBS_BATCH_StatusJob. The read indicates a 
successful response was received. The question now remains why did it 
take so long. Joshua suggested it might be a naming issue. I would look 
there. However, you know your setup the best. I hope this helps.

Ken


More information about the torqueusers mailing list