cloud-crowd 0.2.3 → 0.2.4

data/cloud-crowd.gemspec

@@ -1,7 +1,7 @@
 Gem::Specification.new do |s|
   s.name      = 'cloud-crowd'
-  s.version   = '0.2.3'         # Keep version in sync with cloud-cloud.rb
-  s.date      = '2009-09-23'
+  s.version   = '0.2.4'         # Keep version in sync with cloud-cloud.rb
+  s.date      = '2009-09-28'
 
   s.homepage    = "http://wiki.github.com/documentcloud/cloud-crowd"
   s.summary     = "Parallel Processing for the Rest of Us"

data/lib/cloud-crowd.rb

@@ -44,7 +44,7 @@ module CloudCrowd
   autoload :WorkUnit,     'cloud_crowd/models'
   
   # Keep this version in sync with the gemspec.
-  VERSION        = '0.2.3'
+  VERSION        = '0.2.4'
   
   # Increment the schema version when there's a backwards incompatible change.
   SCHEMA_VERSION = 3
@@ -162,6 +162,9 @@ module CloudCrowd
         @actions[name] = Module.const_get(Inflector.camelize(name))
       end
       @actions
+    rescue NameError => e
+      adjusted_message = "One of your actions failed to load. Please ensure that the name of your action class can be deduced from the name of the file. ex: 'word_count.rb' => 'WordCount'\n#{e.message}" 
+      raise NameError.new(adjusted_message, e.name)
     end
   end
   

data/lib/cloud_crowd/models/job.rb

@@ -35,7 +35,7 @@ module CloudCrowd
       return queue_for_workers(outs) if merging?
       if complete?
         update_attributes(:outputs => outs, :time => time_taken)
-        fire_callback if callback_url
+        Thread.new { fire_callback } if callback_url
       end
       self
     end

data/lib/cloud_crowd/models/node_record.rb

@@ -39,11 +39,12 @@ module CloudCrowd
       result = node['/work'].post(:work_unit => unit.to_json)
       unit.assign_to(self, JSON.parse(result)['pid'])
       touch && true
-    rescue Errno::ECONNREFUSED # Couldn't post to node, assume it's gone away.
-      destroy && false
     rescue RestClient::RequestFailed => e
       raise e unless e.http_code == 503 && e.http_body == Node::OVERLOADED_MESSAGE
       update_attribute(:busy, true) && false
+    rescue RestClient::Exception, Errno::ECONNREFUSED 
+      # Couldn't post to node, assume it's gone away.
+      destroy && false
     end
     
     # What Actions is this Node able to run?

data/lib/cloud_crowd/models/work_unit.rb

@@ -7,6 +7,10 @@ module CloudCrowd
   class WorkUnit < ActiveRecord::Base
     include ModelStatus
     
+    # We use a random number in (0...MAX_RESERVATION) to reserve work units.
+    # The size of the maximum signed integer in MySQL -- SQLite has no limit.
+    MAX_RESERVATION = 2147483647
+    
     belongs_to :job
     belongs_to :node_record
     
@@ -15,7 +19,9 @@ module CloudCrowd
     # Available WorkUnits are waiting to be distributed to Nodes for processing.
     named_scope :available, {:conditions => {:reservation => nil, :worker_pid => nil, :status => INCOMPLETE}}
     # Reserved WorkUnits have been marked for distribution by a central server process.
-    named_scope :reserved,  {:conditions => {:reservation => $$}, :order => 'updated_at asc'}
+    named_scope :reserved,  lambda {|reservation_number| 
+      {:conditions => {:reservation => reservation_number}, :order => 'updated_at asc'}
+    }
     
     # Attempt to send a list of WorkUnits to nodes with available capacity.
     # A single central server process stops the same WorkUnit from being
@@ -27,8 +33,8 @@ module CloudCrowd
     # from the availability list when they are successfully sent, and Nodes get
     # removed when they are busy or have the action in question disabled.
     def self.distribute_to_nodes
-      return unless WorkUnit.reserve_available
-      work_units = WorkUnit.reserved
+      return unless reservation_number = WorkUnit.reserve_available
+      work_units = WorkUnit.reserved(reservation_number)
       available_nodes = NodeRecord.available
       while node = available_nodes.shift and unit = work_units.shift do
         if node.actions.include? unit.action
@@ -40,18 +46,20 @@ module CloudCrowd
         work_units.push(unit)
       end
     ensure
-      WorkUnit.cancel_reservations
+      WorkUnit.cancel_reservations(reservation_number)
     end
     
     # Reserves all available WorkUnits for this process. Returns false if there 
     # were none available.
     def self.reserve_available
-      WorkUnit.available.update_all("reservation = #{$$}") > 0
+      reservation_number = ActiveSupport::SecureRandom.random_number(MAX_RESERVATION)
+      any = WorkUnit.available.update_all("reservation = #{reservation_number}") > 0
+      any && reservation_number
     end
     
     # Cancels all outstanding WorkUnit reservations for this process.
-    def self.cancel_reservations
-      WorkUnit.reserved.update_all('reservation = null')
+    def self.cancel_reservations(reservation_number)
+      WorkUnit.reserved(reservation_number).update_all('reservation = null')
     end
     
     # Cancels all outstanding WorkUnit reservations for all processes. (Useful

data/lib/cloud_crowd/node.rb

@@ -24,6 +24,9 @@ module CloudCrowd
     # (if configured to do so in config.yml).
     MONITOR_INTERVAL    = 3
     
+    # The interval at which the node regularly checks in with central (5 min).
+    CHECK_IN_INTERVAL   = 300
+    
     # The response sent back when this node is overloaded.
     OVERLOADED_MESSAGE  = 'Node Overloaded'
     
@@ -86,6 +89,7 @@ module CloudCrowd
       asset_store
       @server_thread   = Thread.new { @server.start }
       check_in(true)
+      check_in_periodically
       monitor_system if @max_load || @min_memory
       @server_thread.join
     end
@@ -100,7 +104,7 @@ module CloudCrowd
         :max_workers      => CloudCrowd.config[:max_workers],
         :enabled_actions  => @enabled_actions.join(',')
       )
-    rescue Errno::ECONNREFUSED
+    rescue RestClient::Exception, Errno::ECONNREFUSED
       puts "Failed to connect to the central server (#{@central.to_s})."
       raise SystemExit if critical
     end
@@ -160,6 +164,18 @@ module CloudCrowd
       end
     end
     
+    # If communication is interrupted for external reasons, the central server 
+    # will assume that the node has gone down. Checking in will let central know 
+    # it's still online.
+    def check_in_periodically
+      @check_in_thread = Thread.new do
+        loop do
+          sleep CHECK_IN_INTERVAL
+          check_in
+        end
+      end
+    end
+    
     # Trap exit signals in order to shut down cleanly.
     def trap_signals
       Signal.trap('QUIT') { shut_down }
@@ -170,6 +186,7 @@ module CloudCrowd
     
     # At shut down, de-register with the central server before exiting.
     def shut_down
+      @check_in_thread.kill if @check_in_thread
       @monitor_thread.kill if @monitor_thread
       check_out
       @server_thread.kill if @server_thread

data/test/config/actions/failure_testing.rb

@@ -2,7 +2,7 @@
 # all of its retries.
 class FailureTesting < CloudCrowd::Action
   
-  def run
+  def process
     if options['attempts'] + 1 >= CloudCrowd.config[:work_unit_retries]
       return 'made it!'
     else

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification 
 name: cloud-crowd
 version: !ruby/object:Gem::Version 
-  version: 0.2.3
+  version: 0.2.4
 platform: ruby
 authors: 
 - Jeremy Ashkenas
@@ -9,7 +9,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2009-09-23 00:00:00 -04:00
+date: 2009-09-28 00:00:00 -04:00
 default_executable: 
 dependencies: 
 - !ruby/object:Gem::Dependency