OpenPBS Job State

The job state E is too long to finish, Can some ways improve it?

Please check the mom logs on the compute node that ran this job.($PBS_HOME/mom_logs)
It seems there is an issue with passwordless-ssh (scp) between the compute node and server host for that user.

Yeah,I have set the passwordless-ssh, it works. The E would occasionally appear, but it disappears quickly. Thanks for your help, my friend. By the way, Can you tell me the reason of that? And how can I see state if a job is abnormal or fault when it’s running?
Be deeply grateful.

“E” job is in existing state doing the stage out (copying results from compute node to server node)

And how can I see state if a job is abnormal or fault when it’s running?

PBS Pro is like a postman it does not inspect your job or find out what is happening with respect to your job(application solving a problem). It waits for the exist status and based on that job has succeeded or failed. Either you have to tail the log files when it is running the job ( please check this No output and error file generated run time - #2 by adarsh )

qstat -fx | grep -i exit_status.

I have created output and error file, why doesn’t it store in local file?
What kinds of results or information or message will it copy from compute node to server node?What files will the computer node copy results to?How will the server node use the results?

Please refer this section 3.3.6 Writing Files Directly to Final Destination
from this document

Oh, I get it, thank you.