User Tools

Site Tools


cluster:managing_jobs

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
cluster:managing_jobs [2021/05/03 02:40] abraslavcluster:managing_jobs [2024/11/11 20:30] (current) – external edit 127.0.0.1
Line 1: Line 1:
 ======Managing Jobs====== ======Managing Jobs======
-=====Managing Jobs on the Cluster===== 
-The above commands work well when you have not logged out of the cluster while running the job.  Because of the cluster's [[Hardware_Configuration|setup]], there are many nodes that could be running your job.  When you log in to the cluster, you are automatically sent to a particular node <code>econ7</code> through <code>econ18</code> This might not be the same node that is running your job. 
  
-One way to get to the node that has your job is log in and then go to the particular node like: 
- <code>ssh econ3</code> 
- 
-And then to see processes that are running under your username in this node: 
- <code>ps -ux</code> 
- 
-There are also special cluster tools to manage jobs between nodes.  You can use the command 
- <code>cluster-ps</code> 
- 
-This returns a list of processes on each node that are associated with your user id. (As of 2014-06-26 it only works for econ0-5.)  You can choose a process to kill by noting the node and process id ('PID' in the cluster-ps output) and then using the command 
- <code>cluster-kill -n <node> -p <PID></code> 
- 
-======Renice-ing===== 
-You can change how many CPU cycles are given to your job by using the command <code>renice</code> Once you know the process id, you can use the command: 
- <code>renice -n <priority> -p <process id></code> 
-to change the [[nice]]ness of your process.  If you are running a computationally intensive job, please set your job to a high niceness level, so that you don't monopolize cluster resources :) 
 =====General Job Management===== =====General Job Management=====
  
-Once you have a job running, you might want to change it'priority (<code>renice</code>or stop it.  To check on the list of running jobs, use the command <code>top</code>+Once you have a job running, you might want to change its priority or stop it.  To check on the list of running jobs, use the command <code>htop</code>  
- +
-[[http://unixhelp.ed.ac.uk/CGI/man-cgi?top|Top]] is an interactive program that dynamically lists processes running on the server.  Pressing '?' will get you to a help page.  Pressing 'q' gets out of the help page, and also quits top altogether.  +
  
 You can also see all running jobs statically with You can also see all running jobs statically with
Line 36: Line 16:
 Alternatively, press 'q' to stop top.  Then, type use the <code>kill</code> command. Alternatively, press 'q' to stop top.  Then, type use the <code>kill</code> command.
  <code>kill <process id></code>  <code>kill <process id></code>
 +
 +======Renice-ing=====
 +You can change how many CPU cycles are given to your job by using the command <code>renice</code> Once you know the process id, you can use the command:
 + <code>renice -n <priority> -p <process id></code>
 +to change the niceness of your process.  If you are running a computationally intensive job, please set your job to a high niceness level, so that you don't monopolize cluster resources :)
  
cluster/managing_jobs.1620009647.txt.gz · Last modified: 2021/05/03 02:40 (external edit)