Cluster commands that are actually interesting

Try the command “top” on one of the nodes or the SUN servers, it stands for “Table of Processes” and looks something like below, If you notice it is a continuous display, to quit just type “ctrl+c” or the letter “q” and its quits the program, but the interesting thing to note is the username on the left, and how much of the CPU is he/she using on the right. You can do a lot of neat things with this utility, try typing the letter ‘h’ to get a help listing and play around with the few options like sorting the display based on memory usage etc.

In the directory /usr/public/scripts on the master node, you can find some scripts that can be very useful for a cluster user. Look at the script “Runall.pl”, copy this file to your home directory and run it, it gives you a “>” now any command you type will be executed on all the nodes (the nodes which have not crashed) sequentially, so for example you want to check the load average (usage) of the cluster, run the command “uptime” to get an idea of the load on the cluster. You can type “exit” to get out of the program. The script killall.pl is another useful utility that will clean up any crashed programs from all the nodes, use if after a failed job. There are other scripts in there, take a look!