Another scheduling question (using DBMS_JOBS)

Question

Let's say I have a procedure do_task(text IN VARCHAR2, retry_count IN PLS_INTEGER)
I want to schedule a job to run this procedure every N minutes, each time decreasing retry_count by one. When the procedure succeeds in its task or is run with retry_count=0, the job should be killed.
So initially I might do:
DECLARE
string VARCHAR2;
job_no NUMBER;
BEGIN
string:= 'do_task(' || text || ', 8);'
DBMS_JOBS.submit(job_no,string,SYSDATE,'SYSDATE+5/(24*60)');
COMMIT;
END;
Does all this seem normal? The other way is to have a database table recording this stuff but I don't want tables if not required. And how can a job know its own id to adjust itself?

Answer 1

Actually I think I can instead use DBA_JOBS.FAILURES to see the number of retries. I think this is incremented each time the job becomes broken? The thing I can't find is how the scheduler decides whether a job has become broken? If my job's execution fails can I prevent the job being broken (so it doesn't stop being repeated), and manually update the FAILURES field?