[ANN] slave-0.2.0

ruby-talk
1

SYNOPSIS

   the Slave class forks a process and starts a drb server in the child using
   any object as the server. the process is detached so it is not required
   (nor possible) to wait on the child pid. a Heartbeat is set up between the
   parent and child processes so that the child will exit of the parent exits
   for any reason - preventing orphaned slaves from running indefinitely. the
   purpose of Slaves is to be able to easily set up a collection of objects
   communicating via drb protocols instead of having to use IPC.

   typical usage:

     obj = AnyClass::new

     slave = Slave::new obj

     p slave.object # handle on drb object
     p slave.uri # uri of the drb object
     p slave.socket # unix domain socket path for drb object
     p slave.psname # title shown in ps/top

   other usage:

   set the pulse_rate used for the Heartbeat

     slave = Slave::new MyClass::new, 'pulse_rate' => 10

   same

     Slave::pulse_rate = 10
     slave = Slave::new MyClass::new

   same

     ENV['SLAVE_PULSE_RATE'] = 10
     slave = Slave::new MyClass::new

   to avoid having a copy of the object in both the parent and child use the
   block form

     slave = Slave::new{ Server::new } # copy only in child!
     server = slave.object

   if both an object AND a block are passed the object is passed to the block
   in the child process

     slave = Slave::new(Server::new){|server| p 'in child!' }

   slaves may be configured via the environment, the Slave class, or via the
   ctor for object itself. attributes which may be configured include

     * socket_creation_attempts
     * pulse_rate
     * psname
     * debug

HISTORY

   0.2.0:
     incorporated joel vanderWerf's patch such that, if no object is passed the
     block is used to create one ONLY in the child. this avoids having a copy
     in both parent and child is that needs to be avoided due to, for instance,
     resource consumption.

   0.0.1:
     - patch from Logan Capaldo adds block form to slave new, block is run in the
       child

     - added a few more samples/*

     - added Slave#wait

     - added status information to slaves

     - added close-on-exec flag to pipes in parent process

   0.0.0:
     - initial version

enjoy.

-a

···

--
in order to be effective truth must penetrate like an arrow - and that is
likely to hurt. -- wei wu wei

2

This is a great library Ara, I am writing a library that wraps up the slave lib into another interface for controlling jobs with timers and a few other things. So the master process holds a thread pool with references to slave classes in it so that you can set a limit on the number of running slaves. Using it to offload jobs from a rails app to not hold up the http flow. I have a question though. What would be the cleanest way to have a slave class signal that it is finished working and needs to be kill'ed. I know there is the shutdown method that you could call from the master process to shutdown a slave. But is there something similar that can be called from a slave process? Or should I just let the slave process exit! and let the heartbeat cleanup after the dead slaves?

Thanks
-Ezra

···

On Sep 23, 2006, at 3:52 PM, ara.t.howard@noaa.gov wrote:

SYNOPSIS

  the Slave class forks a process and starts a drb server in the child using
  any object as the server. the process is detached so it is not required
  (nor possible) to wait on the child pid. a Heartbeat is set up between the
  parent and child processes so that the child will exit of the parent exits
  for any reason - preventing orphaned slaves from running indefinitely. the
  purpose of Slaves is to be able to easily set up a collection of objects
  communicating via drb protocols instead of having to use IPC.

<snip>

-a

3

hmmm. good question. i could add a feature where at_exit handlers could be
added, somthing like

   s = Slave.new( :at_exit => lambda{}){ do_child_stuff }

would that be sufficient?

-a

···

On Sun, 24 Sep 2006, Ezra Zygmuntowicz wrote:

  This is a great library Ara, I am writing a library that wraps up the
  slave lib into another interface for controlling jobs with timers and
  a few other things. So the master process holds a thread pool with
  references to slave classes in it so that you can set a limit on the
  number of running slaves. Using it to offload jobs from a rails app to
  not hold up the http flow. I have a question though. What would be the
  cleanest way to have a slave class signal that it is finished working
  and needs to be kill'ed. I know there is the shutdown method that you
  could call from the master process to shutdown a slave. But is there
  something similar that can be called from a slave process? Or should I
  just let the slave process exit! and let the heartbeat cleanup after
  the dead slaves?

--
in order to be effective truth must penetrate like an arrow - and that is
likely to hurt. -- wei wu wei

4

Ara-

  Yeah that would work fine I think. So that would call the lambda when the child process calls exit! ? Is there any way for a slave to call back to the parent process? Or is there a way to get the drb uri of the parent from the child?

  Thanks for slave and all your other codeforpeople :wink: I have learned a lot by reading through your code and by googling "cat a.rb" :wink:

Cheers-
-Ezra

···

On Sep 24, 2006, at 8:57 AM, ara.t.howard@noaa.gov wrote:

On Sun, 24 Sep 2006, Ezra Zygmuntowicz wrote:

  This is a great library Ara, I am writing a library that wraps up the
  slave lib into another interface for controlling jobs with timers and
  a few other things. So the master process holds a thread pool with
  references to slave classes in it so that you can set a limit on the
  number of running slaves. Using it to offload jobs from a rails app to
  not hold up the http flow. I have a question though. What would be the
  cleanest way to have a slave class signal that it is finished working
  and needs to be kill'ed. I know there is the shutdown method that you
  could call from the master process to shutdown a slave. But is there
  something similar that can be called from a slave process? Or should I
  just let the slave process exit! and let the heartbeat cleanup after
  the dead slaves?

hmmm. good question. i could add a feature where at_exit handlers could be
added, somthing like

  s = Slave.new( :at_exit => lambda{}){ do_child_stuff }

would that be sufficient?

-a