Computations can fail if required resources are unavailable. A common example is a group-based rating computation that uses any “Stage” value of a given increment (e.g. hourly) as input. There may be Stage values in your database for which you do not currently have a rating table.
Recall that computations are triggered by “tasklist” records when an input to a computation is written to the database. You can determine what happens to a tasklist record a when a computation fails:
- Discard the tasklist record. That is, never retry failed computations.
- Retry every hour a limited number of times.
- Retry every hour indefinitely.
In some databases (CWMS) the processing required to support failed-computation-retries has been seen to be prohibitively expensive. We recommend for CWMS that you disable failed computation retries by setting the variable in your “decodes.properties” file:
If you do want to attempt to retry computations, set this to true and then set an additional variable:
Set this to the maximum number of times you want a computation attempted. The special value of zero means to retry indefinitely.