OWB Workflow Help

Question

Im conducting a proof of concept using OWB Workflow version 2.4. To give some background, the objective is to run numerous dependant SQL scripts that in effect hand craft the ETL process, at a pre-scheduled time, that provide logging and re-starting functionality.
With the distinct lack of publications and documentations Id appreciate any feedback anyone may have on the following:
1) Is it possible to restart a process at a step midway through the flow. For example the 3rd SQL Activity step out of 10 may have failed and instead of starting over, its a requirement to re-start the step that failed.
2) I've read that Oracle are rumoured to be de-supporting the workflow functionality, if so then what is their proposed method of scheduling and controlling ETL jobs.
3) Can a log file generated by the SQL Plus activity be attached to the failure or success email activity?
Any input would be appreciated, also any pointers to OWB publications or documentation, other than the Oracle manuals of course.
Thanks in advance
James

Answer 1

We have been working towards something similar, and the approach we have settled on is as follows:
1) We have added our own runtime-audit tables which track the most recent process flow run date, final status, and the completion status of each mapping. So, a new run can see if a) this run is a restart, in which case we set our max date for drawing data over to be the run date of the last run, and b) what mappings ran succesfully. This way, a function called needs_to_run(mapping_name varchar2) returns a true/false answer to see if a given mapping needs to be executed on this run. The logic is pretty simple
if run_type is full_run or (run_type is restart and nvl(last_run_status,'FAILURE') of this mapping is 'FAILURE')
then
return true;
else
return false;
2) We do not use mapping objects in the process flows. We will only use SQL*Plus objects that do something like this:
variable exec_return_code;
declare
retVal varchar2(30) := 'OK';
begin
if needs_to_run('ThisMapping') then
ThisMapping.main(retVal);
end if;
:exec_return_code := retVal;
exception
when others then
:exec_return_code := 'FAILURE';
end;
exit :exec_return_code
So that each mapping will check first to see if it needs to execute before running thus skipping over fully completed mappings in the case of a restart. As to emailing log information, we currently do that in our post-mapping procedures in the event of a failure rather than logging to file and then attaching to email. We also have a last step that emails a summary of our audit info for each mapping run - but again this is all done inside pl/sql rather than spooling to file.
Anyway, just some ideas to toss your way. And no, we haven't yet got it all completed and working yet, but it is coming along nicely and none of our individual proof-of-concept tests have provided us with any show-stoppers thus far.
Cheers,
Mike