Description
Currently, task is added to the queue whenever we receive an IActiveContext event. In fault tolerant case, as evaluators can be failed, and we might submit a new evaluator and context, and we need to wait until receiving all the active contexts then start to submit tasks.
The TaskSubmitor is responsible for:
- Creating a new Communication group
- Preparing task configuration
- Add tasks to TaskManager
- Submit Tasks
The current code need to be refactored to support it.
Attachments
Issue Links
- Is contained by
-
REEF-1223 IMRU Fault Tolerance - restart failed evaluators
- Resolved