Once you are logged into a cluster (fortyfour.ibest.uidaho.edu), you are on a machine which is referred to as the headnode or the master. This is a server used as the central hub for the cluster. It contains the everything needed to submit, view and check your jobs on the cluster. Please be aware that running applications on the headnode is forbidden and will be removed without warning unless prior arrangements have been made. We are rather draconian in our handling of these events and utilize a wonderful tool called killall, which will not only kill the process in question but everything including terminal connections that the user is running. Generally you will receive a warning message before your session is disconnected that will go something like this:
jcamel@fortyfour ~ $ ./my_big_program Message from firstname.lastname@example.org on pts/9 at 17:20 ... Please do not run applications on the headnode EOF Connection to fortyfour.ibest.uidaho.edu closed by remote host. Connection to fortyfour.ibest.uidaho.edu closed.
Multiple violations will result in the account being suspended.
The reason why we do not allow jobs to be run on the headnode, is due to serious problems we have had in the past. Some of our users have spawned multiple applications, due to clusters being full or not understanding how to use the cluster, which have used up all the resources on the headnode. Once these resources have been used by user processes, it leaves none for the processes which control the cluster. Once in a while we have had user spawned processes which have brought down the head node entirely. This behavior can not only affect you, but will likely affect the jobs of many others. As a result, we do not allow anything, without prior approval, to be run on the headnode If you have questions regarding this or would like approval to run specific applications on the headnode, contact the RCDS Systems Administrators. Better yet, just use one of our many standalone servers - where you can run jobs directly.