gistfile1.rb

# Job 
  add_column :current_batch_job_id    # current batch job running

# This represents batch job runs - and provides history
create_table "batch_jobs", :force => true do |t|
  t.integer   "job_id"                # job this is related to
  t.string    "state"                 # state: create, processed, running, failed, stopped, finished
  t.integer   "instances_to_create"   # count of the planned number of instances to create
  t.datetime  "created_at"
  t.datetime  "updated_at"
end

# This represents sections of a job that need to be run and provides history
create_table "batch_job_instances", :force => true do |t|
  t.integer   "batch_job_id"          # batch job this is part of
  t.string    "state"                 # new, assigned_to_ec2, running, failed, stopped, finished
  t.string    "current_ec2_instance_id"       # refernce to current ec2_instance 
  t.text      "row_ids"               # all ids to process
  t.text      "ids_processing"        # ids waiting to be processed
  t.text      "ids_failed"            # ids that failed to be processed
  t.text      "ids_processed"         # ids finished processing
  t.datetime  "created_at"
  t.datetime  "updated_at"
end

# This represents actual ec2-instances
create_table "ec2_instances", :force => true do |t|
  t.integer   "batch_job_instance_id"
  t.string    "aws_ec2_instance_id"
  t.string    "state"                 # building, running, failed, deleted
  t.datetime  "created_at"
  t.datetime  "updated_at"
end
  
Scenario:  Running a job
1) create batch_job row
2) create batch_job_instaces

Out of process:  something running sees a batch_job_instance.state = new
1) in a transaction - grabs row, changes state to assigned_to_ec2
2) creates ec2 instance - creates ec2_instance row with information
3) does whatever makes sense for state machine in ec2_instance row
4) when everythings running changes state of batch_job_instance to running