Severity: wishlist
thanks
While you raise a valid concern, this is not a bug in distributed-net.
It is a feature of the Linux kernel.
Distributed.net runs with a nice level on 19. This is the lowest
possible priority that a process can run at on a UNIX system. So in
theory while any other application is running the scheduler should
choose it over distributed.net.
While, in theory, this would work that way that you expect it to,
reality is slightly more complex. If your process makes a blocking
system call (usually to perform an I/O operation) then the kernel is
free to perform a context switch and allow distributed.net to run until
some time after the call that caused your process to block completes and
your process becomes runable. This could explain why you are seeing
distributed-net running 5% of the time.
The effect of having distributed-net running in the background should be
somewhere between zero and negligible, especially for short jobs (less
then a few hours). If you are concerned that distributed.net is using
too much cpu time that should go to other tasks then you have to either
shut it down, or suspend it. Suspending it will tell the scheduler to
not run it at all (even if the cpu is idle), but keep it in memory. You
can suspend distributed-net by running the following as root:
kill -STOP <PID of distributed-net>
and to resume:
kill -CONT <PID of distributed-net>
calling:
/usr/bin/distributed-net -pause and /usr/bin/distributed-net -unpause
should work as well.
While this is not a bug in distributed-net, your concern is valid, so I
will look at updating the documentation and/or adding support for
suspending distributed-net to the init script.
James
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
