[GE users] Solution to qdel not killing processes
margaret_Doll at brown.edu
Fri Mar 13 13:35:07 GMT 2009
I have a Rock 5.3 and RedHat 2.6.18-53.1.14.el. "qdel" was deleting jobs from the queue manager; the jobs were not showing up in any of the qstat commands or in qmon, but the processes were still running on the compute nodes.
A friend of mine gave me a solution:
libdshconfig-0.20.13 was installed in /usr/local using
Then dsh-0.25.9 was installed in /usr/local using
A list of machines had to be installed in /usr/local/etc/machines.list. The compute nodes which allow queued jobs were included in the file.
/usr/local/etc/dsh.conf was edited to change "remoteshell=rsh" to "remoteshell=ssh". rsh is not a secure process.
qdel-all was installed /opt/gridengine/bin/lx26-amd64 where the other q commands are located. It contains
if($#argv != 2) then
echo 2 arguments are required, first the process number followed by the executable
/usr/local/dsh-0.25.9/dsh -a "killall -9 $2"
The user runs "qdel-all queueProcess# jobName" An example
qdel-all 20492 qserch-1
More information about the gridengine-users