god 0.6.0 → 0.7.0

data/History.txt

@@ -1,4 +1,70 @@
-== 0.6.0 / 
+== 0.7.0 / 2008-02-01
+  * Minor Enhancements
+    * Better default pid_file_directory behavior
+    * Add --attach <pid> to specify that god should quit if <pid> exits
+  * Bug Fixes
+    * Handle ECONNRESET in HttpResponseCode
+
+== 0.6.12 / 2008-01-31
+  * Minor Enhancements
+    * Allow log file output for non-daemonized god
+    * Switch to SIGTERM from SIGHUP for default lambda killer
+
+== 0.6.11 / 2008-01-31
+  * Major Enhancements
+    * HUGE refactor of timer system to simplify scheduling
+  * Minor Enhancements
+    * Check for a truly working event system and disallow event conditions if none is present
+
+== 0.6.10 / 2008-01-24
+  * Bug Fixes
+    * Fix ensure_stop nil pid no local variable bug
+
+== 0.6.9 / 2008-01-23
+  * Bug Fixes
+    * Fix Timer condition dedup behavior
+
+== 0.6.8 / 2008-01-23
+  * Minor Enhancements
+    * Warn if a command returns a non-zero exit code
+    * Ensure that stop command actually stops process
+
+== 0.6.7 / 2008-01-22
+  * Minor Enhancements
+    * Add --no-syslog option to disable Syslog
+    * Allow contact redeclaration (dups are ignored)
+
+== 0.6.6 / 2008-01-07
+  * Bug Fixes
+    * Redo Timer mutexing to reduce synchronization needs
+
+== 0.6.5 / 2008-01-04
+  * Bug Fixes
+    * Fix Timer descheduling deadlock issue
+    * Change HttpResponseCode to use GET instead of HEAD
+
+== 0.6.4 / 2008-12-31
+  * Bug Fixes
+    * Refactor Hub to clarify mutexing
+    * Eliminate potential iteration problem in Timer
+    * Add caching PID accessor to process to solve event deregistration failure
+
+== 0.6.3 / 2007-12-18
+  * Minor Enhancements
+    * Output ProcessExits registration/deregistration info
+
+== 0.6.2 / 2007-12-17
+  * Minor Enhancements
+    * Output registered PID for ProcessExits
+  * Bug Fixes
+    * Fix `god remove <group>` not working for unmonitored watches
+
+== 0.6.1 / 2007-12-14
+
+* Minor Enhancement
+  * Log when state change is complete
+
+== 0.6.0 / 2007-12-4
 
 * Minor Enhancement
   * Move Syslog calls into God::Logger and clean up all calling code

data/Manifest.txt

@@ -34,12 +34,13 @@ lib/god/configurable.rb
 lib/god/contact.rb
 lib/god/contacts/email.rb
 lib/god/dependency_graph.rb
+lib/god/diagnostics.rb
+lib/god/driver.rb
 lib/god/errors.rb
 lib/god/event_handler.rb
 lib/god/event_handlers/dummy_handler.rb
 lib/god/event_handlers/kqueue_handler.rb
 lib/god/event_handlers/netlink_handler.rb
-lib/god/hub.rb
 lib/god/logger.rb
 lib/god/metric.rb
 lib/god/process.rb
@@ -49,7 +50,6 @@ lib/god/sugar.rb
 lib/god/system/process.rb
 lib/god/task.rb
 lib/god/timeline.rb
-lib/god/timer.rb
 lib/god/trigger.rb
 lib/god/watch.rb
 test/configs/child_events/child_events.god
@@ -84,10 +84,10 @@ test/test_conditions_process_running.rb
 test/test_conditions_tries.rb
 test/test_contact.rb
 test/test_dependency_graph.rb
+test/test_driver.rb
 test/test_event_handler.rb
 test/test_god.rb
 test/test_handlers_kqueue_handler.rb
-test/test_hub.rb
 test/test_logger.rb
 test/test_metric.rb
 test/test_process.rb
@@ -97,6 +97,5 @@ test/test_sugar.rb
 test/test_system_process.rb
 test/test_task.rb
 test/test_timeline.rb
-test/test_timer.rb
 test/test_trigger.rb
 test/test_watch.rb

data/Rakefile

@@ -1,7 +1,7 @@
 require 'rubygems'
 require 'hoe'
 
-Hoe.new('god', '0.6.0') do |p|
+Hoe.new('god', '0.7.0') do |p|
   p.rubyforge_name = 'god'
   p.author = 'Tom Preston-Werner'
   p.email = 'tom@rubyisawesome.com'

data/bin/god

@@ -9,7 +9,7 @@ require 'optparse'
 require 'drb'
 
 begin
-  options = {:daemonize => true, :port => 17165}
+  options = {:daemonize => true, :port => 17165, :syslog => true}
 
   opts = OptionParser.new do |opts|
     opts.banner = <<-EOF
@@ -71,10 +71,28 @@ begin
     opts.on("-V", "Print extended version and build information") do
       options[:info] = true
     end
+    
+    opts.on("--log-level LEVEL", "Log level [debug|info|warn|error|fatal]") do |x|
+      options[:log_level] = x.to_sym
+    end
+    
+    opts.on("--no-syslog", "Disable output to syslog") do
+      options[:syslog] = false
+    end
+    
+    opts.on("--attach PID", "Quit god when the attached process dies") do |x|
+      options[:attach] = x
+    end
   end
   
   opts.parse!
   
+  # validate
+  if options[:log_level] && ![:debug, :info, :warn, :error, :fatal].include?(options[:log_level])
+    abort("Invalid log level '#{options[:log_level]}'")
+  end
+  
+  # dispatch
   if !options[:config] && options[:version]
     require 'god'
     God::CLI::Version.version

data/lib/god.rb

@@ -6,9 +6,16 @@ require 'rubygems'
 # core
 require 'stringio'
 require 'logger'
+require 'fileutils'
+
+begin
+  require 'fastthread'
+rescue LoadError
+ensure
+  require 'thread'
+end
 
 # stdlib
-require 'syslog'
 
 # internal requires
 require 'god/errors'
@@ -42,8 +49,7 @@ require 'god/contact'
 require 'god/contacts/email'
 
 require 'god/socket'
-require 'god/timer'
-require 'god/hub'
+require 'god/driver'
 
 require 'god/metric'
 require 'god/watch'
@@ -58,6 +64,8 @@ require 'god/sugar'
 require 'god/cli/version'
 require 'god/cli/command'
 
+require 'god/diagnostics'
+
 $:.unshift File.join(File.dirname(__FILE__), *%w[.. ext god])
 
 # App wide logging system
@@ -76,13 +84,6 @@ $run ||= nil
 
 GOD_ROOT = File.expand_path(File.join(File.dirname(__FILE__), '..'))
 
-# Ensure that Syslog is open
-begin
-  Syslog.open('god')
-rescue RuntimeError
-  Syslog.reopen('god')
-end
-
 # Return the binding of god's root level
 def root_binding
   binding
@@ -132,12 +133,13 @@ class Module
 end
 
 module God
-  VERSION = '0.6.0'
+  VERSION = '0.7.0'
   
-  LOG_BUFFER_SIZE_DEFAULT = 1000
-  PID_FILE_DIRECTORY_DEFAULT = '/var/run/god'
+  LOG_BUFFER_SIZE_DEFAULT = 10
+  PID_FILE_DIRECTORY_DEFAULTS = ['/var/run/god', '~/.god/pids']
   DRB_PORT_DEFAULT = 17165
   DRB_ALLOW_DEFAULT = ['127.0.0.1']
+  LOG_LEVEL_DEFAULT = :info
   
   class << self
     # user configurable
@@ -146,7 +148,8 @@ module God
                        :port,
                        :allow,
                        :log_buffer_size,
-                       :pid_file_directory
+                       :pid_file_directory,
+                       :log_level
     
     # internal
     attr_accessor :inited,
@@ -157,7 +160,8 @@ module God
                   :watches,
                   :groups,
                   :contacts,
-                  :contact_groups
+                  :contact_groups,
+                  :main
   end
   
   # initialize class instance variables
@@ -167,6 +171,7 @@ module God
   self.allow = nil
   self.log_buffer_size = nil
   self.pid_file_directory = nil
+  self.log_level = nil
   
   # Initialize internal data.
   #
@@ -185,10 +190,20 @@ module God
     
     # set defaults
     self.log_buffer_size ||= LOG_BUFFER_SIZE_DEFAULT
-    self.pid_file_directory ||= PID_FILE_DIRECTORY_DEFAULT
     self.port ||= DRB_PORT_DEFAULT
     self.allow ||= DRB_ALLOW_DEFAULT
-    LOG.level = Logger::INFO
+    self.log_level ||= LOG_LEVEL_DEFAULT
+    
+    # additional setup
+    self.setup
+    
+    # log level
+    log_level_map = {:debug => Logger::DEBUG,
+                     :info => Logger::INFO,
+                     :warn => Logger::WARN,
+                     :error => Logger::ERROR,
+                     :fatal => Logger::FATAL}
+    LOG.level = log_level_map[self.log_level]
     
     # init has been executed
     self.inited = true
@@ -290,6 +305,8 @@ module God
     if watch.group
       self.groups[watch.group].delete(watch)
     end
+    
+    applog(watch, :info, "#{watch.name} unwatched")
   end
   
   # Instantiate a new Contact of the given kind and send it to the block.
@@ -324,9 +341,10 @@ module God
       self.uncontact(existing_contact)
     end
     
-    # ensure the new contact has a unique name
+    # warn and noop if the contact has been defined before
     if self.contacts[c.name] || self.contact_groups[c.name]
-      abort "Contact name '#{c.name}' already used for a Contact or Contact Group"
+      applog(nil, :warn, "Contact name '#{c.name}' already used for a Contact or Contact Group")
+      return
     end
     
     # abort if the Contact is invalid, the Contact will have printed
@@ -373,30 +391,30 @@ module God
   #
   # Returns String[]:task_names
   def self.control(name, command)
-    # get the list of watches
-    watches = Array(self.watches[name] || self.groups[name])
+    # get the list of items
+    items = Array(self.watches[name] || self.groups[name]).dup
     
     jobs = []
     
     # do the command
     case command
       when "start", "monitor"
-        watches.each { |w| jobs << Thread.new { w.monitor if w.state != :up } }
+        items.each { |w| jobs << Thread.new { w.monitor if w.state != :up } }
       when "restart"
-        watches.each { |w| jobs << Thread.new { w.move(:restart) } }
+        items.each { |w| jobs << Thread.new { w.move(:restart) } }
       when "stop"
-        watches.each { |w| jobs << Thread.new { w.unmonitor.action(:stop) if w.state != :unmonitored } }
+        items.each { |w| jobs << Thread.new { w.unmonitor.action(:stop) if w.state != :unmonitored } }
       when "unmonitor"
-        watches.each { |w| jobs << Thread.new { w.unmonitor if w.state != :unmonitored } }
+        items.each { |w| jobs << Thread.new { w.unmonitor if w.state != :unmonitored } }
       when "remove"
-        watches.each { |w| jobs << Thread.new { self.unwatch(w) } }
+        items.each { |w| self.unwatch(w) }
       else
         raise InvalidCommandError.new
     end
     
     jobs.each { |j| j.join }
     
-    watches.map { |x| x.name }
+    items.map { |x| x.name }
   end
   
   # Unmonitor and stop all tasks.
@@ -513,20 +531,40 @@ module God
   end
   
   def self.setup
-    # Make pid directory
-    unless test(?d, self.pid_file_directory)
-      begin
-        FileUtils.mkdir_p(self.pid_file_directory)
-      rescue Errno::EACCES => e
-        abort "Failed to create pid file directory: #{e.message}"
+    if self.pid_file_directory
+      # pid file dir was specified, ensure it is created and writable
+      unless File.exist?(self.pid_file_directory)
+        begin
+          FileUtils.mkdir_p(self.pid_file_directory)
+        rescue Errno::EACCES => e
+          abort "Failed to create pid file directory: #{e.message}"
+        end
+      end
+      
+      unless File.writable?(self.pid_file_directory)
+        abort "The pid file directory (#{self.pid_file_directory}) is not writable by #{Etc.getlogin}"
+      end
+    else
+      # no pid file dir specified, try defaults
+      PID_FILE_DIRECTORY_DEFAULTS.each do |idir|
+        dir = File.expand_path(idir)
+        begin
+          FileUtils.mkdir_p(dir)
+          if File.writable?(dir)
+            self.pid_file_directory = dir
+            break
+          end
+        rescue Errno::EACCES => e
+        end
+      end
+      
+      unless self.pid_file_directory
+        dirs = PID_FILE_DIRECTORY_DEFAULTS.map { |x| File.expand_path(x) }
+        abort "No pid file directory exists, could be created, or is writable at any of #{dirs.join(', ')}"
       end
     end
-  end
-  
-  def self.validater
-    unless test(?w, self.pid_file_directory)
-      abort "The pid file directory (#{self.pid_file_directory}) is not writable by #{Etc.getlogin}"
-    end
+    
+    applog(nil, :info, "Using pid file directory: #{self.pid_file_directory}")
   end
   
   # Initialize and startup the machinery that makes god work.
@@ -534,18 +572,10 @@ module God
   # Returns nothing
   def self.start
     self.internal_init
-    self.setup
-    self.validater
     
     # instantiate server
     self.server = Socket.new(self.port)
     
-    # start event handler system
-    EventHandler.start if EventHandler.loaded?
-    
-    # start the timer system
-    Timer.get
-    
     # start monitoring any watches set to autostart
     self.watches.values.each { |w| w.monitor if w.autostart? }
     
@@ -555,8 +585,15 @@ module God
     # mark as running
     self.running = true
     
-    # join the timer thread so we don't exit
-    Timer.get.join
+    # don't exit
+    self.main = 
+    Thread.new do
+      loop do
+        sleep 60
+      end
+    end
+    
+    self.main.join
   end
   
   # To be called on program exit to start god

data/lib/god/cli/command.rb

@@ -77,6 +77,12 @@ module God
         begin
           Signal.trap('INT') { exit }
           name = @args[1]
+          
+          unless name
+            puts "You must specify a Task or Group name"
+            exit!
+          end
+          
           t = Time.at(0)
           loop do
             print @server.running_log(name, t)
@@ -141,7 +147,7 @@ module God
             
             God::EventHandler.register(pid, :proc_exit) do
               puts "[ok] process exit event received"
-              exit(0)
+              exit!(0)
             end
             
             sleep(1)

data/lib/god/cli/run.rb

@@ -12,6 +12,7 @@ module God
         # have at_exit start god
         $run = true
         
+        # run
         if @options[:daemonize]
           run_daemonized
         else
@@ -19,6 +20,19 @@ module God
         end
       end
       
+      def attach
+        process = System::Process.new(@options[:attach])
+        Thread.new do
+          loop do
+            unless process.exists?
+              applog(nil, :info, "Going down because attached process #{@options[:attach]} exited")
+              exit!
+            end
+            sleep 5
+          end
+        end
+      end
+      
       def run_daemonized
         # trap and ignore SIGHUP
         Signal.trap('HUP') {}
@@ -40,6 +54,11 @@ module God
               puts
             end
             
+            # start attached pid watcher if necessary
+            if @options[:attach]
+              self.attach
+            end
+            
             # set port if requested
             if @options[:port]
               God.port = @options[:port]
@@ -50,12 +69,26 @@ module God
               God.pid = @options[:pid] 
             end
             
+            unless @options[:syslog]
+              Logger.syslog = false
+            end
+            
             # load config
             if @options[:config]
+              # set log level, defaults to WARN
+              if @options[:log_level]
+                God.log_level = @options[:log_level]
+              else
+                God.log_level = :warn
+              end
+              
               unless File.exist?(@options[:config])
                 abort "File not found: #{@options[:config]}"
               end
               
+              # start the event handler
+              God::EventHandler.start if God::EventHandler.loaded?
+              
               begin
                 load File.expand_path(@options[:config])
               rescue Exception => e
@@ -73,6 +106,7 @@ module God
             STDIN.reopen "/dev/null"
             STDOUT.reopen(log_file, "a")
             STDERR.reopen STDOUT
+            STDOUT.sync = true
           rescue => e
             puts e.message
             puts e.backtrace.join("\n")
@@ -92,15 +126,28 @@ module God
       def run_in_front
         require 'god'
         
+        # start attached pid watcher if necessary
+        if @options[:attach]
+          self.attach
+        end
+        
         if @options[:port]
           God.port = @options[:port]
         end
         
+        # set log level if requested
+        if @options[:log_level]
+          God.log_level = @options[:log_level]
+        end
+        
         if @options[:config]
           unless File.exist?(@options[:config])
             abort "File not found: #{@options[:config]}"
           end
           
+          # start the event handler
+          God::EventHandler.start if God::EventHandler.loaded?
+          
           begin
             load File.expand_path(@options[:config])
           rescue Exception => e
@@ -112,6 +159,17 @@ module God
               abort "There was an error in your configuration file (see above)"
             end
           end
+          
+          if @options[:log]
+            log_file = File.expand_path(@options[:log])
+            puts "Sending output to log file: #{log_file}"
+            
+            # reset file descriptors
+            STDIN.reopen "/dev/null"
+            STDOUT.reopen(log_file, "a")
+            STDERR.reopen STDOUT
+            STDOUT.sync = true
+          end
         end
       end
     end # Run