I want to host a Ray server application in a Docker image. Following the example in Customer Docker Images, I created a Python script called fake.py.
from faker import Faker
from ray import serve
@serve.deployment
def create_fake_email():
return Faker().email()
app = create_fake_email.bind()
In the same directory I have a Dockerfile.
FROM rayproject/ray:latest
RUN pip install Faker==18.13.0
WORKDIR /serve_app
COPY fake.py /serve_app/fake.py
EXPOSE 8000 8265
ENTRYPOINT ["serve", "run", "fake:app"]
I build the image.
docker build . -t ray-server
Then I run the image.
docker run -p 8000:8000 -p 8265:8265 ray-server
The ray server starts. From inside the container I can connect to the dashboard on port 8265 and run the server application on port 8000, but neither of these work from the host machine.
curl http://localhost:8000
curl: (7) Failed to connect to localhost port 8000 after 0 ms: Couldn't connect to server
curl http://localhost:8265
curl: (7) Failed to connect to localhost port 8265 after 0 ms: Couldn't connect to server
I’m sure I’m doing all the Docker commands correctly. This appears to be the same as a problem reported for Ray 2.5.0 in the About Dashboard inside Docker post on this board.
Is this a bug? Does anyone have a workaround?
How severely does this issue affect your experience of using Ray?
- High: It blocks me from completing my task.