This page shows how the cluster is being used.<p/>

Every 5 minutes, a script running on ippc11 (as the ipp user. See its
crontab) starts a process on each node of the cluster. There, five
(i.e. the number of iterations) snapshots of the node activity are
made at an interval of 0.5 second (i.e. time interval).<br/>

Statistics about processes that have been <b>observed at least
twice</b> are then computed. They are displayed if at least one of
these conditions is satisfied (the corresponding column is then shown
in red):
<ul>
  <li>The owner of the process is owned by an "unexpected" user
  (i.e. by someone else
  than <tt>root</tt>, <tt>nobody</tt>, <tt>iqanalysis</tt>...)</li>
  <li>The owner of the process is <tt>ipp</tt> or <tt>ippdvo</tt>
  (even if they are "expected" users), mainly because we want to check
  if anything is still running while, e.g., we are changing the
  tag.</li>
  <li>The process used more than 5% of the total CPU on average</li>
  <li>The process used more than 5% of the total memory on average</li>
  <li>The process is older than 6 hours</li>
</ul>

The displayed columns are the following:
<ul>
  <li>The process ID (i.e. pid) with its number of observations</li>
  <li>The process owner</li>
  <li>The process cpu usage (a multi-threaded process may use more than 100%)</li>
  <li>The process memory usage</li>
  <li>The process age in hours</li>
  <li>The process command line</li>
</ul>

Feel free to correct this documentation and to send me your suggestions and comments.<br/>

Serge