I’m trying to get a ray serve application working on an internal company cluster.
I started on my dev laptop using a kind cluster.
I followed this example:
https://docs.ray.io/en/master/serve/production-guide/kubernetes.html#installing-the-kuberay-operator
to the letter.
It works. Great.
Next, I downloaded
https://github.com/ray-project/test_dag/archive/41d09119cbdf8450599f993f51318e9e27c59098.zip
which contains fruit.py, the example.
I uploaded the exact same zip file to the company’s artifactory repo (which I cannot share).
If I run wget on the linux cli I can download the zip file.
I changed the file ray_v1alpha1_rayservice.yaml
runtimeEnv: |
working_dir: “the github link”
to the link to my company’s artifactory repo, instead of the github link.
Example doesn’t work. It spins up the ray cluster, but
it will not create rayservice-sample-head-svc or rayservice-sample-serve-svc, the latter of which is how you get responses.
Is there an issue with artifactory?
Is there another way to get the main code I want to run working?
Can I put fruit.py in a docker container, load that docker container onto all the workers and the head and then reference the containerized fruit.py in the cluster yaml?