Rerun a step in a chain

Question

i want to rerun a step after it was marked as failed.
if step 1 failed notify the dba and invoke step 1 again else invoke step 2.
i wrote the folowing :
DBMS_SCHEDULER.DEFINE_CHAIN_STEP(
chain_name => 'reports.some_chain',
step_name => 'step_1' ,
program_name => 'do_something1');
DBMS_SCHEDULER.DEFINE_CHAIN_STEP(
chain_name => 'reports.some_chain',
step_name => 'step_2' ,
program_name => 'do_something_else');
DBMS_SCHEDULER.DEFINE_CHAIN_STEP(
chain_name => 'reports.some_chain',
step_name => 'step_fail' ,
program_name => 'notify_dba');
DBMS_SCHEDULER.DEFINE_CHAIN_RULE (
chain_name => 'reports.some_chain',
condition => 'step_1 SUCCEEDED',
action => 'START step_2',
rule_name => 'step_1_success'
DBMS_SCHEDULER.DEFINE_CHAIN_RULE (
chain_name => 'reports.some_chain',
condition => 'step_2 SUCCEEDED',
action => 'END',
rule_name => 'step_2_success'
DBMS_SCHEDULER.DEFINE_CHAIN_RULE (
chain_name => 'reports.some_chain',
condition => 'step_1 FAILED',
action => 'START step_fail',
rule_name => 'step_1_failed'
DBMS_SCHEDULER.DEFINE_CHAIN_RULE (
chain_name => 'reports.some_chain',
condition => 'step_fail COMPLETED',
action => 'START step_1',
rule_name => 'step_retry_step_1'
this works fine if step 1 ends successfuly. when step 1 failes and "step_fail" is triggered, step 1 is not being executed again. i cant understand why.
i would also like to know how to use the syntax for 'AFTER X SECONDS START step'.
thanks !

Answer 1

Hi,
The reason this doesn't work is that once step1 has is in a completed state the rules will not let it run again.
There is a way to get around this though. In your step_fail program you can set step1 to be not_started yet or you can set step1 to start immediately by directly modifying the running chain e.g.
exec dbms_scheduler.alter_running_chain('chainjob1','step1', 'state', 'RUNNING')
or
exec dbms_scheduler.alter_running_chain('chainjob1','step1', 'state', 'NOT_STARTED')
both of these will run step1 again.
Your notify_dba program can be modified to take the step_name as one of the arguments (define_metadata_argument job subname) so it will know which step failed.
Hope this helps,
Ravi.