Details
-
Improvement
-
Status: Open
-
Major
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
When scheduling a non-trivial Trogdor workoad (e.g 15 agents and 150 tasks), it is difficult to diagnose all the tasks that are running.
The output of `GET /coordinator/tasks` can get very large - 28.6MB in my case. This is mainly due to running a lot of ProduceBenchSpec tasks which are subscribed to a lot of topics.
I think we should have a simpler endpoint which exposes all the task IDs that are running. This can let us get the IDs to query specific tasks via `GET /coordinator/tasks?taskId=<id1>,<id2>` ot `GET /coordinator/tasks/<task_id>`