Do not retry failed jobs by default
In case of a failure, by default each job is retried 2 times without any delay in between.
This usually makes no sense as the circumstances do not change automatically after a failure. E.g. an authorization exception will likely also occur 2nd and 3rd time after the first one. A connection exception usually also does not resolve within one second automatically. Additionally the customer usually gets charged for AI services called within a task so that we do not want that to be re-triggered in an uncontrolled manner.
This has changed now:
-
The past behavior was:
-
Failed jobs are retried 2 times (3 times at all)
-
-
The new behavior is:
-
Failed jobs should not be retried
-
This comes with the following benefits:
-
Lowering the risk of calling paid services multiple times in general
-
Lowering the complexity for new WFLs: The default should be "no retry at all" so that WFL devs do not have to set this value deliberately. If a retry makes sense then WFL devs have to define that retry strategy.