Backgroundrb devel - Nov 2008 - Stopping bdrb from another process gets the process killed

Hi, I''ve been having some issues integrating bdrb on a cruisecontrolrb
server
It''s related to the way we stop the bdrb server. The thing is, right
now
when we stop the process we are killing the parent of the group of the
process. This is ok if you''re running script/backgroundrb stop directly
from
the console, but if I''m running this from another process,
let''s say a rake
task, the rake task gets killed because it gets the SIGTERM signal.

In the class StarStop we have:

  def kill_process arg_pid_file
    pid = nil
    pid = File.open(arg_pid_file, "r") { |pid_handle|
pid_handle.gets.strip.chomp.to_i }
    pgid =  Process.getpgid(pid)
    puts "Stopping BackgrounDRb with pid #{pid}...."
    Process.kill(''-TERM'', pgid)
    File.delete(arg_pid_file) if File.exists?(arg_pid_file)
    puts "Success!"
  end

What I did was, that instead of killing pgid, I''m doing a system call
and
killing the pid, like this

  def kill_process arg_pid_file
    pid = nil
    pid = File.open(arg_pid_file, "r") { |pid_handle|
pid_handle.gets.strip.chomp.to_i }
    puts "Stopping BackgrounDRb with pid #{pid}...."
    system("kill -s TERM #{pid}") #This is what''s different
    File.delete(arg_pid_file) if File.exists?(arg_pid_file)
    puts "Success!"
  end

I did the system call because Process, says that there''s no process
with the
pid we extract from the file.

It''s working good, but I don''t know if there are some hidden
side effects?
any ideas?

here''s a rake task so you can try what I''m saying
http://pastie.org/318949
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://rubyforge.org/pipermail/backgroundrb-devel/attachments/20081120/cc8da325/attachment.html>

On Thu, Nov 20, 2008 at 4:18 PM, Jorge Dias <theturingmachine at
gmail.com> wrote:
> Hi, I''ve been having some issues integrating bdrb on a
cruisecontrolrb
> server
> It''s related to the way we stop the bdrb server. The thing is,
right now
> when we stop the process we are killing the parent of the group of the
> process. This is ok if you''re running script/backgroundrb stop
directly from
> the console, but if I''m running this from another process,
let''s say a rake
> task, the rake task gets killed because it gets the SIGTERM signal.
>
> In the class StarStop we have:
>
>   def kill_process arg_pid_file
>     pid = nil
>     pid = File.open(arg_pid_file, "r") { |pid_handle|
> pid_handle.gets.strip.chomp.to_i }
>     pgid =  Process.getpgid(pid)
>     puts "Stopping BackgrounDRb with pid #{pid}...."
>     Process.kill(''-TERM'', pgid)
>     File.delete(arg_pid_file) if File.exists?(arg_pid_file)
>     puts "Success!"
>   end
>
> What I did was, that instead of killing pgid, I''m doing a system
call and
> killing the pid, like this
>
>   def kill_process arg_pid_file
>     pid = nil
>     pid = File.open(arg_pid_file, "r") { |pid_handle|
> pid_handle.gets.strip.chomp.to_i }
>     puts "Stopping BackgrounDRb with pid #{pid}...."
>     system("kill -s TERM #{pid}") #This is what''s
different
>     File.delete(arg_pid_file) if File.exists?(arg_pid_file)
>     puts "Success!"
>   end
>
> I did the system call because Process, says that there''s no
process with the
> pid we extract from the file.
>
> It''s working good, but I don''t know if there are some
hidden side effects?
> any ideas?
>
> here''s a rake task so you can try what I''m saying
http://pastie.org/318949
Hmm, I will play with this and let you know. Doesn''t look like much of
problem, except that system call is opaque.