Looking for a method to kill all children for my Timeout::Error contingency


#1

Do I really need to collect all the pids from fork to kill them in that
case, or
is there a method I’m not finding in pickaxe that just does it???

xc


#2

On Wed, Mar 25, 2009 at 04:57:03AM +0900, Xeno C. wrote:

Do I really need to collect all the pids from fork to kill them in that case, or
is there a method I’m not finding in pickaxe that just does it???

Here’s a trick I use, maybe it is useful. Note that even if
push-to-slaves
creates children, those too will be killed because they reside in the
same
process group.

require ‘timeout’
signal = ‘-TERM’
timeout = 30

pid = Process.fork do
Process.setsid
exec %{push-to-slaves -v}
end
begin
Timeout::timeout(timeout) { Process.waitpid(pid) }
rescue Timeout::Error => exc
puts “#{exc.message}, killing pid #{pid}”
Process.kill(signal, pid)
res = Process.waitpid(pid)
puts “pid #{pid.inspect} != res #{res.inspect}” if res != pid
end


#3

Jos B. wrote:

On Wed, Mar 25, 2009 at 04:57:03AM +0900, Xeno C. wrote:

Do I really need to collect all the pids from fork to kill them in that case, or
is there a method I’m not finding in pickaxe that just does it???

Here’s a trick I use, maybe it is useful. Note that even if push-to-slaves
creates children, those too will be killed because they reside in the same
process group.

The problem with that is I have many processes, so I would be doing this
with
arrays and loops, which suddenly becomes problematic. If I could just
call
something that automatically kills the entire tree of children on down
that
would be much nicer…??


#4

On Wed, Mar 25, 2009 at 09:56:53AM +0900, Mike B wrote:

The problem with that is I have many processes, so I would be doing this with
arrays and loops, which suddenly becomes problematic. If I could just call
something that automatically kills the entire tree of children on down that
would be much nicer…??

If those many processes were being started by the push-to-slaves script
as in
the example I posted, unless that script did something to create new
process
groups for its children, the code I posted would terminate all those
processes
because they all belong to the same process group which is targeted by
the
Process.kill(-TERM, pid) invocation. Iow, if you want to kill all the
children
at once the easiest way is to make sure they all belong to the same
pgrp.


#5

On Mar 24, 12:23 pm, Xeno C. removed_email_address@domain.invalid wrote:

arrays and loops, which suddenly becomes problematic. If I could just call

puts "pid #{pid.inspect} != res #{res.inspect}" if res != pid

end


“Facts do not cease to exist because they are ignored” - Aldous Huxley
“…Or I might add, because they are denied” - Dr. Albert A. Bartletthttp://www.eskimo.com/~xeno

Maybe this could help?

http://markmail.org/message/3nl2idghbekgbahb

It’s a link to Ara Howards announcement for terminator.

Mike B.


#6

Jos B. wrote:

On Wed, Mar 25, 2009 at 09:56:53AM +0900, Mike B wrote:

On Mar 24, 12:23 pm, Xeno C. removed_email_address@domain.invalid wrote:

Jos B. wrote:

Yes thank you.

Part of my problem is I wasn’t feeling sure about how I would use
processes or
threads, and why. With the advent of the new Chrome Philosophy, I was
thinking
perhaps processes might be the way to go, as I was doing curl
executions, but it
turns out I need to do a surrounding sequence of things in parallel for
initializations and closures, so I am leaning back to threads now, which
I think
solves the problems. I would sure like to find a book just on use of
threads
and processes in Ruby. I would think a knowledgeable person could
probably write
at least 12 chapters on it for the layman alone, and not having done
some of
this for over ten years, that’s where I’m feeling I’m at.

Thank you again for the feedback.

Sincerely, Xeno

xc