Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
Description
At present, the RPC communication from the master to an executor happens in 2 places: sending a task, and replying a block location request. We should properly handle exceptions that occur when these RPCs fail due to reasons such as executor failure.
For now, we can choose to simply ignore NCS RPC failures assuming that executor failures will be handled by the FailedEvaluator event.
Attachments
Issue Links
- is blocked by
-
NEMO-50 Carefully retry tasks in the scheduler
- Resolved
- links to