[Bug] Ray Mac OSX security popups every time I run script locally

[worldveil]

Search before asking

• I searched the issues and found no similar issues.

Ray Component

Ray Core, Monitoring & Debugging, Others

What happened + What you expected to happen

I always see these popups.

They are ridiculously annoying and often keep appearing even after script is done running.

It doesn't give me confidence in Ray.

Reproduction script

Literally any ray example will work.

from ray import tune


def objective(step, alpha, beta):
    return (0.1 + alpha * step / 100)**(-1) + beta * 0.1


def training_function(config):
    # Hyperparameters
    alpha, beta = config["alpha"], config["beta"]
    for step in range(10):
        # Iterative training function - can be any arbitrary training procedure.
        intermediate_score = objective(step, alpha, beta)
        # Feed the score back back to Tune.
        tune.report(mean_loss=intermediate_score)


analysis = tune.run(
    training_function,
    config={
        "alpha": tune.grid_search([0.001, 0.01, 0.1]),
        "beta": tune.choice([1, 2, 3])
    })

print("Best config: ", analysis.get_best_config(
    metric="mean_loss", mode="min"))

# Get a dataframe for analyzing trial results.
df = analysis.results_df

should trigger this.

Anything else

Let's figure out how to address and add to the docs.

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

[ericl]

@ijrsvt @thomasdesr or other mac users, do you know if this is because the worker processes are binding to something like 0.0.0.0 (instead of 127.0.0.1)?

Or is OSX just complaining about any port being opened, even if it listens only internally?

[pcmoritz]

Did we try if listening on localhost on macOS fixes this issue as @ericl suggests above? That should fix this issue right? Is this hard to do?

[jjyao]

I did a quick check, for grpc servers, we are listening to 0.0.0.0 and gcs server is using 192.168. Do we currently have a way to know in c++ that we are in the single node mode so we can safely use 127.0.0.1. Once we know that, the fix should be trivial. @ericl @pcmoritz?

[pcmoritz]

I don't think there is a flag/setting for single node mode currently, but I think it would be ok to add a simple helper function in C++ that checks if we are on macOS and if yes return 127.0.0.1 and else return 0.0.0.0. Later if needed we could migrate it to a flag, but I don't really see the need. I doubt we need to support macOS clusters any time soon (or ever).

[worldveil]

I can still repro this:

# requirements.txt
ray[default,data,tune] @ https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-2.0.0.dev0-cp37-cp37m-macosx_10_15_intel.whl
jupyter
torch
awscli
pyarrow  # for loadng parquet
scikit-learn
tensorboard
boto3
fsspec
pickle5

# dask ecosystem
dask
dask_ml
s3fs

# mlflow
mlflow
torchvision

then run:

import ray
import time

ray.init() 

@ray.remote
def test():
    time.sleep(10)
    print("hello!")
    return 10

results = ray.get([test.remote() for i in range(10)])

[simon-mo]

can you confirm the import ray; print(ray.__commit__)?

[jjyao]

I can reproduce it with a new conda environment. But I'm not able to reproduce with locally built ray with the same commit.

[jjyao]

Also If I just do pip install -U https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-2.0.0.dev0-cp37-cp37m-macosx_10_15_intel.whl and run the script. There is also no popups.

[worldveil]

@simon-mo do you need me to check the commit still?

thank you all for digging in here!

[jjyao]

@worldveil I don't need the commit and I have a PR out for the fix. Will let you verify once it's merged.

[jjyao]

@worldveil it should have been fixed. Could you verify and close the issue?

[worldveil]

@jjyao I still see this on ray start --head

[jjyao]

@worldveil Can you try it now with the latest wheel. Sorry for so many iterations.

[jjyao]

Confirmed with @worldveil that popups are gone.

[worldveil]

@jjyao I have seen it again :/ is there any reason I should be seeing this on Ray 1.9.x? any regressions?

I can try to provide a repro if needed, but it is inconsistent

[jjyao]

No, it shouldn't happen. If you can give a repro, I'm happy to debug it.

[worldveil]

Have not been able to repro! Leaving closed for now.