Oozie: run at some time or at some frequency, whichever comes first -
the benefits of coordinating absolute time (insofar jobs take consistent amount of time) output ready others @ time (e.g. update dashboard during night people see in morning).
the benefits of coordinating relative frequency that, if oozie (or it's server) down, no jobs skipped (e.g. daily job might run 2 hours late, not 22 hours late).
how can like:
start="2009-01-01t21:00z" frequency="${coord:days(1)}" run-if-skipped="true"
i.e. when all's well, jobs run daily @ 9pm. if happens oozie (e.g. server rolled) between 8pm , 10pm, once oozie comes @ 10pm, job should run @ 10pm, , tomorrow @ 9pm normal.
not sure understand question.
if server down, , re-start coordinator, start coordinator start time.
also can make job run every hour, check if output folder exist - stop . use decision control node
Comments
Post a Comment