We are observing quite some task-activities that are failing with TypeMissing exception.
This results in backoff and reduces the performance significantly.
This becomes interesting, especially when the type (TaskActivity/TaskOrchestration) is intentionally removed assuming old ones would die eventually. (Especially for the case of testing).
We can't delete them manually, as message is read by taskhubworker and is hidden.
There are cases where orchestrator is already completed, while task-activity is still trying to run.
Ask:
Can there be a way, where one can provide the ETA for type, implies, when it is picked up for processing, if the time it queued was older than a threshold interval, just mark it finished.
For task-activities for which orchestration is already completed, can those just be removed without processing?
A way to kill one type of type (say orchestration of type A, just remove all those entries from the queue, similarly for task-activity and other work-items). Like some API etc. E.g., AddNoOpTaskOrchestration, AddNoOpTaskActivity.
Problem statement:
Ask: