Class: CloudCrowd::Job
- Inherits:
-
ActiveRecord::Base
- Object
- ActiveRecord::Base
- CloudCrowd::Job
- Includes:
- ModelStatus
- Defined in:
- lib/cloud_crowd/models/job.rb
Overview
A chunk of work that will be farmed out into many WorkUnits to be processed in parallel by each active CloudCrowd::Worker. Jobs are defined by a list of inputs (usually public urls to files), an action (the name of a script that CloudCrowd knows how to run), and, eventually a corresponding list of output.
Constant Summary collapse
- CLEANUP_GRACE_PERIOD =
That’s a week.
7
Class Method Summary collapse
-
.cleanup_all(opts = {}) ⇒ Object
Clean up all jobs beyond a certain age.
-
.create_from_request(h) ⇒ Object
Create a Job from an incoming JSON request, and add it to the queue.
Instance Method Summary collapse
-
#action_class ⇒ Object
Retrieve the class for this Job’s Action.
-
#all_work_units_complete? ⇒ Boolean
Have all of the WorkUnits finished?.
-
#any_work_units_failed? ⇒ Boolean
Have any of the WorkUnits failed?.
-
#check_for_completion ⇒ Object
After work units are marked successful, we check to see if all of them have finished, if so, continue on to the next phase of the job.
-
#cleanup_assets ⇒ Object
Cleaning up after a job will remove all of its files from S3 or the filesystem.
-
#color ⇒ Object
Generate a stable 8-bit Hex color code, based on the Job’s id.
-
#done_splitting? ⇒ Boolean
This job is done splitting if it’s finished with its splitting work units.
-
#fire_callback ⇒ Object
If a
callback_url
is defined, post the Job’s JSON to it upon completion. -
#mergeable? ⇒ Boolean
This job is mergeable if its Action has a
merge
method. -
#percent_complete ⇒ Object
How complete is this Job? Unfortunately, with the current processing sequence, the percent_complete can pull a fast one and go backwards.
-
#set_next_status ⇒ Object
Transition this Job’s current status to the appropriate next one, based on the state of the WorkUnits and the nature of the Action.
-
#splittable? ⇒ Boolean
This job is splittable if its Action has a
split
method. -
#time_taken ⇒ Object
How long has this Job taken?.
-
#to_json(opts = {}) ⇒ Object
A JSON representation of this job includes the statuses of its component WorkUnits, as well as any completed outputs.
Methods included from ModelStatus
#complete?, #display_status, #failed?, included, #incomplete?, #merging?, #processing?, #splitting?, #succeeded?
Class Method Details
.cleanup_all(opts = {}) ⇒ Object
Clean up all jobs beyond a certain age.
35 36 37 38 39 40 |
# File 'lib/cloud_crowd/models/job.rb', line 35 def self.cleanup_all(opts = {}) days = opts[:days] || CLEANUP_GRACE_PERIOD self.complete.older_than(days).find_in_batches(:batch_size => 100) do |jobs| jobs.each {|job| job.destroy } end end |
.create_from_request(h) ⇒ Object
Create a Job from an incoming JSON request, and add it to the queue.
24 25 26 27 28 29 30 31 32 |
# File 'lib/cloud_crowd/models/job.rb', line 24 def self.create_from_request(h) self.create( :inputs => h['inputs'].to_json, :action => h['action'], :options => (h['options'] || {}).to_json, :email => h['email'], :callback_url => h['callback_url'] ) end |
Instance Method Details
#action_class ⇒ Object
Retrieve the class for this Job’s Action.
117 118 119 120 121 |
# File 'lib/cloud_crowd/models/job.rb', line 117 def action_class @action_class ||= CloudCrowd.actions[self.action] return @action_class if @action_class raise Error::ActionNotFound, "no action named: '#{self.action}' could be found" end |
#all_work_units_complete? ⇒ Boolean
Have all of the WorkUnits finished?
92 93 94 |
# File 'lib/cloud_crowd/models/job.rb', line 92 def all_work_units_complete? self.work_units.incomplete.count <= 0 end |
#any_work_units_failed? ⇒ Boolean
Have any of the WorkUnits failed?
97 98 99 |
# File 'lib/cloud_crowd/models/job.rb', line 97 def any_work_units_failed? self.work_units.failed.count > 0 end |
#check_for_completion ⇒ Object
After work units are marked successful, we check to see if all of them have finished, if so, continue on to the next phase of the job.
44 45 46 47 48 49 50 51 52 53 54 |
# File 'lib/cloud_crowd/models/job.rb', line 44 def check_for_completion return unless all_work_units_complete? set_next_status outs = gather_outputs_from_work_units return queue_for_workers([outs]) if merging? if complete? update_attributes(:outputs => outs, :time => time_taken) Thread.new { fire_callback } if callback_url end self end |
#cleanup_assets ⇒ Object
Cleaning up after a job will remove all of its files from S3 or the filesystem. Destroying a Job will cleanup_assets first. Run this in a separate thread to get out of the transaction’s way. TODO: Convert this into a ‘cleanup’ work unit that gets run by a worker.
87 88 89 |
# File 'lib/cloud_crowd/models/job.rb', line 87 def cleanup_assets AssetStore.new.cleanup(self) end |
#color ⇒ Object
Generate a stable 8-bit Hex color code, based on the Job’s id.
143 144 145 |
# File 'lib/cloud_crowd/models/job.rb', line 143 def color @color ||= Digest::MD5.hexdigest(self.id.to_s)[-7...-1] end |
#done_splitting? ⇒ Boolean
This job is done splitting if it’s finished with its splitting work units.
107 108 109 |
# File 'lib/cloud_crowd/models/job.rb', line 107 def done_splitting? splittable? && work_units.splitting.count <= 0 end |
#fire_callback ⇒ Object
If a callback_url
is defined, post the Job’s JSON to it upon completion. The callback_url
may include HTTP basic authentication, if you like:
http://user:[email protected]/job_complete
If the callback URL returns a ‘201 Created’ HTTP status code, CloudCrowd will assume that the resource has been successfully created, and the Job will be cleaned up.
74 75 76 77 78 79 80 81 |
# File 'lib/cloud_crowd/models/job.rb', line 74 def fire_callback begin response = RestClient.post(callback_url, {:job => self.to_json}) Thread.new { self.destroy } if response && response.code == 201 rescue RestClient::Exception => e puts "Job ##{id} failed to fire callback: #{callback_url}" end end |
#mergeable? ⇒ Boolean
This job is mergeable if its Action has a merge
method.
112 113 114 |
# File 'lib/cloud_crowd/models/job.rb', line 112 def mergeable? self.processing? && self.action_class.public_instance_methods.map {|m| m.to_sym }.include?(:merge) end |
#percent_complete ⇒ Object
How complete is this Job? Unfortunately, with the current processing sequence, the percent_complete can pull a fast one and go backwards. This happens when there’s a single large input that takes a long time to split, and when it finally does it creates a whole swarm of work units. This seems unavoidable.
128 129 130 131 132 133 134 |
# File 'lib/cloud_crowd/models/job.rb', line 128 def percent_complete return 99 if merging? return 100 if complete? unit_count = work_units.count return 100 if unit_count <= 0 (work_units.complete.count / unit_count.to_f * 100).round end |
#set_next_status ⇒ Object
Transition this Job’s current status to the appropriate next one, based on the state of the WorkUnits and the nature of the Action.
58 59 60 61 62 63 64 65 |
# File 'lib/cloud_crowd/models/job.rb', line 58 def set_next_status update_attribute(:status, any_work_units_failed? ? FAILED : self.splitting? ? PROCESSING : self.mergeable? ? MERGING : SUCCEEDED ) end |
#splittable? ⇒ Boolean
This job is splittable if its Action has a split
method.
102 103 104 |
# File 'lib/cloud_crowd/models/job.rb', line 102 def splittable? self.action_class.public_instance_methods.map {|m| m.to_sym }.include? :split end |
#time_taken ⇒ Object
How long has this Job taken?
137 138 139 140 |
# File 'lib/cloud_crowd/models/job.rb', line 137 def time_taken return self.time if self.time Time.now - self.created_at end |
#to_json(opts = {}) ⇒ Object
A JSON representation of this job includes the statuses of its component WorkUnits, as well as any completed outputs.
149 150 151 152 153 154 155 156 157 158 159 160 161 |
# File 'lib/cloud_crowd/models/job.rb', line 149 def to_json(opts={}) atts = { 'id' => id, 'color' => color, 'status' => display_status, 'percent_complete' => percent_complete, 'work_units' => work_units.count, 'time_taken' => time_taken } atts['outputs'] = JSON.parse(outputs) if outputs atts['email'] = email if email atts.to_json end |