app.rb in documentcloud-cloud-crowd-0.0.6

- old
+ new

@@ -2,21 +2,22 @@
   
   # The main CloudCrowd (Sinatra) application. The actions are:
   #
   # == Admin
   # [get /] Render the admin console, with a progress meter for running jobs.
-  # [get /jobs] Get the combined JSON of every active job in the queue.
+  # [get /status] Get the combined JSON of every active job and worker.
   # [get /heartbeat] Returns 200 OK to let monitoring tools know the server's up.
   # 
   # == Public API
   # [post /jobs] Begin a new Job. Post with a JSON representation of the job-to-be. (see examples).
   # [get /jobs/:job_id] Check the status of a Job. Response includes output, if the Job has finished.
   # [delete /jobs/:job_id] Clean up a Job when you're done downloading the results. Removes all intermediate files.
   #
   # == Internal Workers API
   # [post /work] Dequeue the next WorkUnit, and hand it off to the worker.
   # [put /work/:unit_id] Mark a finished WorkUnit as completed or failed, with results.
+  # [put /worker] Keep a record of an actively running worker.
   class App < Sinatra::Default
     
     set :root, ROOT
     set :authorization_realm, "CloudCrowd"
     
@@ -33,15 +34,27 @@
     # Render the admin console.
     get '/' do
       erb :index
     end
     
-    # Get the JSON for every active job in the queue.
-    get '/jobs' do
-      json Job.incomplete
+    # Get the JSON for every active job in the queue and every active worker
+    # in the system. This action may get a little worrisome as the system grows
+    # larger -- keep it in mind.
+    get '/status' do
+      json(
+        'jobs'            => Job.incomplete, 
+        'workers'         => WorkerRecord.alive(:order => 'name desc'),
+        'work_unit_count' => WorkUnit.incomplete.count
+      )
     end
     
+    # Get the JSON for a worker record's work unit, if one exists.
+    get '/worker/:name' do
+      record = WorkerRecord.find_by_name params[:name]
+      json((record && record.work_unit) || {})
+    end
+    
     # To monitor the central server with Monit, God, Nagios, or another 
     # monitoring tool, you can hit /heartbeat to make sure.
     get '/heartbeat' do
       "buh-bump"
     end
@@ -64,10 +77,12 @@
     delete '/jobs/:job_id' do
       current_job.cleanup
       json nil
     end
     
+    # INTERNAL WORKER DAEMON API:
+    
     # Internal method for worker daemons to fetch the work unit at the front
     # of the queue. Work unit is marked as taken and handed off to the worker.
     post '/work' do
       json dequeue_work_unit
     end
@@ -86,9 +101,16 @@
           json dequeue_work_unit(1)
         else             
           error(500, "Completing a work unit must specify status.")
         end
       end
+    end
+    
+    # Every so often workers check in to let the central server know that
+    # they're still alive. Keep up-to-date records
+    put '/worker' do
+      params[:terminated] ? WorkerRecord.check_out(params) : WorkerRecord.check_in(params)
+      json nil
     end
     
   end
   
 end
\ No newline at end of file