[CSC 435] Jobs Killed

[Andrew J. Pounds]

Long and Jon,
  
  I noticed both of you were utilizing the cluster improperly this evening.

Jon -- you had an interactive PBS session running that had been idle for 
hours with no sign of processes running on the host node.   In essence 
you had blocked anyone else from using any of the machines in room 306 
for the entire evening.  Your job was deleted from the interactive queue.

Long -- you had processes running on numerous processors in lab 204 
OUTSIDE of PBS.  The processes had been running for almost four hours.  
This drastically impacts performance of anyone trying to utilize the 
systems via PBS.  I manually logged into each of the machines in lab 204 
and killed your processes.

PBS is out there to govern fair usage of the cluster -- please use it 
appropriately.   If you are writing code and trying to test parallel 
execution, use a small number of processors interactively via PBS.  If 
you are doing production runs to test scalability, then use batch mode 
PBS.   During the time that we are in room 306 on TR, I have no problem 
with you logging into individual machines and doing things interactively 
on the cluster.  That is time dedicated explicitly to you for doing 
those things.  But once we hit the 7:00 pm -- 7:00 am window, or a 
weekend, all things running over the cluster have to go through PBS.

"We toys can see EVERYTHING.  So PLAY NICE!"

http://www.youtube.com/watch?v=1eHG5tA9_iE

-- 
Andrew J. Pounds, Ph.D.  (pounds at theochem.mercer.edu)
Associate Professor of Chemistry and Computer Science
Mercer University,  Macon, GA 31207   (478) 301-5627