module Terminator
Terminator
¶ ↑
Synopsis¶ ↑
An external timeout mechanism based on processes and signals. Safe for system calls. Safe for minors. but not very safe for misbehaving, downtrodden zombied out processes.
Description¶ ↑
Terminator
is a solution to the problem of 'how am I meant to kill a system call in Ruby!?'
Ruby (at least MRI) uses green threads to “multitask”. This means that there is really only ever one ruby process running which then splits up it's processor time between all of it's threads internally.
The processor then only has to deal with one ruby process and the ruby process deals with all it's threads. There are pros and cons to this method, but that is not the point of this library.
The point is, that if you make a system call to an external resource from ruby, then the kernel will go and make that call for ruby and NOT COME BACK to ruby until that system call completes or fails. This can take a very long time and is why your feeble attempts at using ruby's internal “Timeout” command has failed miserably at timing out your external web service, database or network connections.
You see, Ruby just doesn't get a chance to do anything as the kernel goes “I'm not going to talk to you again until your system calls complete”. Sort of a no win situation for Ruby.
That's where Terminator
comes in. Like Arnie, he will come back. No matter what, and complete his mission, unless he gets aborted before his timeout, you can trust Terminator
to thoroughly and without remorse, nuke your misbehaving and timing out ruby processes efficiently, and quickly.
How it Works¶ ↑
Basically we create a new terminator ruby process, separate to the existing running ruby process that has a simple command of sleep for x seconds, and then do a process TERM on the PID of the original ruby process that created it.
If your process finishes before the timeout, it will kill the Terminator
first.
So really it is a race of who is going to win?
Word of warning though. Terminator
is not subtle. Don't expect it to split hairs. Trying to give a process that takes about 1 second to complete, a 2 second terminator… well… odds are 50/50 on who is going to make it.
If you have a 1 second process, give it 3 seconds to complete. Arnie doesn't much care for casualties of war.
Another word of warning, if using Terminator
inside a loop, it is possible to exceed your open file limit. I have safely tested looping 1000 times
URIS¶ ↑
Usage¶ ↑
The terminator library is simple to use.
require 'terminator' Terminator.terminate(1) do sleep 4 puts("I will never print") end #=> Terminator::Error: Timeout out after 1s
The above code snippet will raise a Terminator::Error
as the terminator's timeout is 2 seconds and the block will take at least 4 to complete.
You can put error handling in with a simple begin / rescue block:
require 'terminator' begin Terminator.terminate(1) do sleep 4 puts("I will never print") end rescue puts("I got terminated, but rescued myself.") end #=> I got terminated, but rescued myself.
The standard action on termination is to raise a Terminator::Error
, however, this is just an anonymous object that is called, so you can pass your own trap handling by giving the terminator a lambda as an argument.
require 'terminator' custom_trap = lambda { eval("raise(RuntimeError, 'Oops... I failed...')") } Terminator.terminate(:seconds => 1, :trap => custom_trap) do sleep 10 end #=> RuntimeError: (eval):1:in `irb_binding': Oops... I failed...
Constants
- Version
Public Class Methods
# File lib/terminator.rb, line 118 def Terminator.dependencies { 'fattr' => [ 'fattr' , ' >= 2.2' ] , } end
# File lib/terminator.rb, line 110 def Terminator.description "an external timeout mechanism based on processes and signals" end
# File lib/terminator.rb, line 114 def Terminator.license "same as ruby's" end
# File lib/terminator.rb, line 106 def Terminator.version Terminator::Version end
Public Instance Methods
# File lib/terminator.rb, line 203 def error() ::Terminator::Error end
# File lib/terminator.rb, line 193 def getopt key, hash, default = nil return hash.fetch(key) if hash.has_key?(key) key = key.to_s return hash.fetch(key) if hash.has_key?(key) key = key.to_sym return hash.fetch(key) if hash.has_key?(key) default end
# File lib/terminator.rb, line 154 def nuke_terminator(pid) Process.kill("KILL", pid) rescue nil Process.wait(pid) rescue nil end
# File lib/terminator.rb, line 159 def plot_to_kill pid, options = {} seconds = getopt :in, options signal = getopt :with, options send_terminator(pid, seconds) end
# File lib/terminator.rb, line 165 def send_terminator(pid, seconds) process = IO.popen(%[#{ ruby } -e'sleep #{seconds}; Process.kill("#{signal}", #{pid}) rescue nil;'], 'w+') process.pid end
# File lib/terminator.rb, line 170 def temp_file_name "terminator-#{ ppid }-#{ pid }-#{ rand }" end
Terminator.terminate
has two ways you can call it. You can either just specify:
Terminator.terminate(seconds) { code_to_execute }
If you want to pass in the block, please use:
Terminator.terminate(:seconds => seconds, :trap => block) { code_to_execute }
Where block is an anonymous method that gets called when the timeout occurs.
# File lib/terminator.rb, line 133 def terminate options = {}, &block options = { :seconds => Float(options) } unless Hash === options seconds = getopt(:seconds, options) default_trap = lambda{ eval("raise(::Terminator::Error, 'Timeout out after #{ seconds }s')", block.binding) } actual_trap = getopt(:trap, options, default_trap) signal_trap = lambda{|*_| actual_trap.call()} previous_trap = Signal.trap(signal, &signal_trap) terminator_pid = plot_to_kill(pid, :in => seconds, :with => signal) begin block.call ensure nuke_terminator(terminator_pid) Signal.trap(signal, previous_trap) end end
# File lib/terminator.rb, line 204 def version() ::Terminator::Version end