[core] Raylets in local cluster do not log to separate files

[stephanie-wang]

What is the problem?

Ray version and other system information (Python version, TensorFlow version, OS): 0.9dev

When multiple raylets are started locally, it used to be that they would each log to a separate file, e.g., raylet.1.out, raylet.2.out, etc. Now, they all log to raylet.out, which makes it harder to debug.

Reproduction (REQUIRED)

Any checked in pytest that starts a local cluster. For example, RAY_BACKEND_LOG_LEVEL=debug pytest -sv python/ray/tests/test_advanced.py::test_wait_cluster

• I have verified my script runs in a clean environment and reproduces the issue.
• I have verified the issue also occurs with the latest wheels.

[stephanie-wang]

I believe this was caused by #8885, @wuisawesome.

[clarkzinzow]

@stephanie-wang That was another regression I noticed, adding a unique=True argument to get_log_file_names here should do the trick:

raylet_out_name, raylet_err_name = self.get_log_file_names("raylet", unique=True)

[stephanie-wang]

@stephanie-wang That was another regression I noticed, adding a unique=True argument to get_log_file_names here should do the trick:

raylet_out_name, raylet_err_name = self.get_log_file_names("raylet", unique=True)

Ah, nice find!

[wuisawesome]

Nice catch, I may have made this same mistake in a few places. @clarkzinzow do you want to submit a PR for this? If not, I'll try to get around to it in a week or so.

[clarkzinzow]

@wuisawesome Sure I can submit a PR for that, I'm already in this code for another PR so it's not much of a context switch.