[ActorInit] Fix Bug in Actor creation

[ijrsvt]

Why are these changes needed?

• [Ray][xlang]Setting async flag for Python actor actor in Java #28149 calls RayActorError with a str as cause, but this is not an accepted type. This leads to hitting the assertion error in the else case: assert isinstance(cause, ActorDiedErrorContext) on L283.

Related issue number

• Noticed this by adding debug statements to the assertion on L283 to figure out why I was hitting it:

(TemporaryActor pid=1225210)   File "/home/ubuntu/anaconda3/envs/my_package/lib/python3.7/site-packages/ray/exceptions.py", line 283, in __init__
(TemporaryActor pid=1225210)     assert isinstance(cause, ActorDiedErrorContext), "Not actor creation: " + str(type(cause)) + "\n: " + str(cause)
(TemporaryActor pid=1225210) AssertionError: Not actor creation: <class 'str'>
(TemporaryActor pid=1225210) : Failed to create the actor ActorID(064094ca4d85ae3e8fc943340e000000). The failure reason is that you set the async flag, but the actor has no any coroutine function.

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[jovany-wang]

Good catch. But it's better to fix it at this line https://github.com/ray-project/ray/blob/master/python/ray/_raylet.pyx#L778

by passing an ActorDiedErrorContext instead of adding a new type.

[ijrsvt]

My concern with doing this is that the RayActorError will not have actor_init_failed=True.

I tried updating this function to return a RayTaskError instead, but then only the traceback is only printed in the str() call, not the cause.

I could do something hacky like:

raise RayActorError.from_task_error(
    RayTaskError(
        function_name=f"{class_name}.__init__",
        traceback_str=f"Failed to create the actor {core_worker}. "
                        "The failure reason is that you set the async flag, "
                        "but the actor has no any coroutine function.",
        cause=""
    )
)

[jovany-wang]

I don't think we should set actor_init_failed=True because this flag indicates that the actor failed when executing __init__.
For our case, the actor should fail before executing __init__.

CC @rkooo567 for double check.

[ijrsvt]

Oh! Thanks for the clarification @jovany-wang! Will move this to ActorDiedErrorContext

[jovany-wang]

LGTM

[ijrsvt]

Failed tests are flaky/failing on master.