Calendar Booking Time for Processes and Server Usage - statonlab/UTIA_Computational_Resource GitHub Wiki
Once you have completed orientation, you will receive access to the UTIA Computational Resources Scheduling Google Calendar. This will allow you to set schedule usage of either Centaur or Sphinx so that other users will know how much of our resources are currently in use at any given point in time. When you run an intensive process, please add it to the schedule as a courtesy to other users.
Whenever you add a process to the calendar, always add how many cores you will be using to the name. If you have an estimate of how long it will take, I would recommend setting the time of the event from when you began running the job to when you believe it will end. Otherwise, you can make it an "all-day" event.
Other ways to track server usage
While we hope the calendar will be used extensively in scheduling jobs, human error is always possible and a job may not be put on the calendar. We have two options to track current server usage:
htop
Running htop
allows you to visualize core usage and provides a list of all currently running processes. You have the option of killing your own jobs if they are not progressing, but you do not have permission to kill other user's jobs. More information about htop can be found here.
ps -ef
Alternatively, running the ps
command provides a list of processes running on the server. (The command is short for "Process Status.") We like to add the -ef
flag to list all running process in full-format listing:
...
mhuff10 1007249 1 0 08:16 ? 00:00:00 /usr/lib/systemd/systemd --user
root 1007251 2 0 08:16 ? 00:00:00 [kworker/65:0]
mhuff10 1007253 1007249 0 08:16 ? 00:00:00 (sd-pam)
root 1007257 2 0 08:16 ? 00:00:00 [kworker/12:1-cgroup_pidlist_destroy]
root 1007258 2 0 08:16 ? 00:00:00 [kworker/12:2-events]
mhuff10 1007261 1007241 0 08:16 ? 00:00:00 sshd: mhuff10@pts/0
mhuff10 1007262 1007261 0 08:16 pts/0 00:00:00 bash
root 1007359 2 0 08:16 ? 00:00:00 [kworker/64:0-events_freezable_power_]
mhuff10 1007363 1007262 0 08:16 pts/0 00:00:00 ps -ef
This generally produces a large number of root processes. To narrow it down to only your processes of interest, you can use the grep
command to search for a string of interest. This can be your username or the name of the package you are currently running. For example, running ps -ef | grep "mhuff10"
gives me the following result:
root 1007241 2484 0 08:16 ? 00:00:00 sshd: mhuff10 [priv]
mhuff10 1007249 1 0 08:16 ? 00:00:00 /usr/lib/systemd/systemd --user
mhuff10 1007253 1007249 0 08:16 ? 00:00:00 (sd-pam)
mhuff10 1007261 1007241 0 08:16 ? 00:00:00 sshd: mhuff10@pts/0
mhuff10 1007262 1007261 0 08:16 pts/0 00:00:00 bash
mhuff10 1007375 1007262 0 08:17 pts/0 00:00:00 ps -ef
mhuff10 1007376 1007262 0 08:17 pts/0 00:00:00 grep --color=auto mhuff10
While both htop
and ps
are helpful tools, do not use them as replacements for scheduling on the calendar.