Big programs
Some programs take a lot of processing time before they
complete. If you are running such a program, and the results are
not needed urgently, you would probably like it to be executing on
a processor when the system is not busy, and for it to be suspended
when the usage of the machine becomes high, in order not to slow
down more urgent processes. The facility called nice
(so called because you are being considerate to other users) exists
to prioritise a job. If mycommand is
a big program, then
$ nice mycommand &
will run mycommand in the background with low
priority.
If a process is already running and you wish to reduce its
priority, the command renice can be used, but we omit
discussion of this command here - examine the manual page using
man for further information.
When you finish your session on the system, a signal
SIGHUP ('hangup') is sent to all the processes you
created during the session. Most processes, when they receive this
signal, will terminate. If you have a job you wish to continue
running in the background after you have logged off - and this will
probably be true for any big jobs you run - you must make the job
immune to the signal SIGHUP by means of command
nohup. The syntax is nohup followed by a
command, you will probably also wish the command to run in the
background:
$ nohup testfile &
You would only want to run a command via nohup in
the foreground if you were connected to the system using a
communication link that might disconnect you without warning in the
middle of a session. Standard output and the standard error stream
from a job running with nohup are redirected to a file
called nohup.out instead of to your terminal, if they
are not already redirected, and the priority of the job is low (as
with nice). The difference between nice
and nohup is principally that a job run with
nice will terminate when the user who invoked the job
logs off.
|