Scheduler - Event Based Jobs

Question

Hi,
I've been trying to create an event based job dependent on multiple jobs. I tried queuing, but it seems like it can only hand one Job at a time. Also, chains were recommended to me, but I want the Job to run on the dependence on the Job that ran on specific program at the specific time. I currently know that chains can only handle, programs, other chains and/or events.
In my mind chains wouldn't work, because If I wanted to run Job3 dependent on the outcome of Job1 running Program1 and Job2 running Program2, chains wouldn't be able to accomdate that. For it to happen in chains I would have to attach Program1 and Program2 to the new chain Job and that is not exactly what I would want.
Can anyone help/clarify this situation for me. Thank you.
Tony

Answer 1

Hi Tony,
So the requirement is to run a job after 2 prior jobs have completed.
There are two ways I can think of to do this, both using events. In both cases you need to set job A and job B to raise events on completion (succeeded, failed or stopped) and setup an agent for the Scheduler event queue in the schema in which you are working
dbms_scheduler.add_event_queue_subscriber('myagent')
In order to get Job C to start after jobs A and B have completed there are 2 options.
1) Let job C point to a chain and start running whenever job A has completed. The chain will have 2 steps, one event step waiting on B and one that runs your task C. An event step does nothing but wait for a particular event and complete successfully when the event is received. The second step will run when the event step waiting on job B has completed.
So your chain would look something like
begin
dbms_scheduler.create_chain('chain1');
dbms_scheduler.define_chain_event_step('chain1','stepB',
'tab.user_data.object_name = ''JOB_B'' and
tab.user_data.event_type IN(''JOB_SUCCEEDED'',''JOB_FAILED'',''JOB_STOPPED''',
'sys.scheduler$_event_queue,myagent');
dbms_scheduler.define_chain_step('chain1','stepC','finalTaskProg');
dbms_scheduler.define_chain_rule('chain1','true','start stepB');
dbms_scheduler.define_chain_rule('chain1', 'stepB completed', 'start stepC');
dbms_scheduler.define_chain_rule('chain1', 'stepC completed', 'end');
dbms_scheduler.enable('chain1');
end;
And your job would point to the chain and run whenever job_A completes (similar condition and queue_spec). It would keep waiting till job_B runs and then the final task would run and it would complete.
2) The second way is to require job_A to insert a row into a table somewhere. Rule conditions can access table data so you could have job_C have an event condition which checks for a completion event for job_B and checks the table to see whether job_A has completed. Then the code you run should then remove the row in the table
e.g.
queue_spec=>'sys.scheduler$_event_queue,myagent'
event_condition=>'tab.user_data.object_name = ''JOB_B'' and
tab.user_data.event_type IN(''JOB_SUCCEEDED'',''JOB_FAILED'',''JOB_STOPPED''' and
(select count(*) from mytab where col='job_A')>0'
Then when running C do - delete from mytab where col='job_a';
Both of these assume that job_A always completes before job_B but both of these will then run job_C after job_B completes. Modifying either of these so that either job A or B runs first is also possible by having another job that waits on A rather than B.
Hope this helps, if you have any more questions, let me know.
-Ravi