[Bug][RayJob] RayJob with custom head service name

[kevin85421]

Why are these changes needed?

As @mmourafiq pointed out in #1326 (comment), some users prefer a deterministic head service name for RayJob in order to set up a reverse proxy. Although we've made the entire head service configurable, there are some issues that arise when we set the name of the head service:

• Kubernetes Jobs cannot communicate with the Ray head because of an incorrect service name.
• Ray workers fail to connect to the Ray head due to the same naming issue.

Related issue number

Closes #1294

Checks

• I've made sure the tests are passing.
• Testing Strategy
  • Unit tests
  • Manual tests
  • This PR is not tested :(

# Step 0: Build a KubeRay operator image
# Step 1: Install a KubeRay operator
helm install kuberay-operator kuberay/kuberay-operator --version 0.6.0 --set image.repository=controller,image.tag=latest

# Step 2: Create a RayJob with custom head service
# https://gist.github.com/kevin85421/b8e0259467ffdb1b000c4177159f57e9#file-ray-job-custom-head-svc-name-yaml
kubectl apply -f ray-job-custom-head-svc-name.yaml

# Step 3: Check RayJob's status
kubectl get rayjobs.ray.io rayjob-sample -o=jsonpath='{.status.jobStatus}'
# [expected output]: "SUCCEEDED"

[mmourafiq]

This looks good 👍

[architkulkarni]

Looks good to me!

[architkulkarni]

Do you think we should add your manual test YAML to the RayJob sample YAML tests so that the e2e behavior is tested in CI?

[kevin85421]

Do you think we should add your manual test YAML to the RayJob sample YAML tests so that the e2e behavior is tested in CI?

This would be great. Should I wait for #1320 or #1321 to be merged or I can work on the test without these 2 PRs? Thanks!

[architkulkarni]

Do you think we should add your manual test YAML to the RayJob sample YAML tests so that the e2e behavior is tested in CI?

This would be great. Should I wait for #1320 or #1321 to be merged or I can work on the test without these 2 PRs? Thanks!

@kevin85421 I think it's independent and you don't need to wait, I think you would just need to add the yaml to the list of tested yamls in test_sample_rayjob_yamls.py

[kevin85421]

@architkulkarni I have already added a test ray-job.custom-head-svc.yaml for this PR.