Ray on Azure tries to install wrong version

How severe does this issue affect your experience of using Ray?

  • None: Just asking a question out of curiosity

I noticed that following the quick start instructions for running Ray on Azure (here: Cluster YAML Configuration Options — Ray 2.2.0 ) doesn’t work - for some reason, Ray tries to install a nightly version on the Azure nodes, and that fails with the following error:

      [5/7] Initializing command runner
  [6/7] Running setup commands
    (0/4) (which conda && echo ‘eval “$(...
Shared connection to closed.
    (1/4) (conda activate py38_tensorflo...
Shared connection to closed.
    (2/4) which ray || pip install -U “r...
ERROR: ray-3.0.0.dev0-cp38-cp38-manylinux2014_x86_64.whl is not a supported wheel on this platform.
Shared connection to closed.
  New status: update-failed
  SSH command failed.
  Failed to setup head node.

In case anyone else runs into this problem, I was able to fix it by manually specifying the setup commands, or running inside docker.
For manual setup commands, just add this to your config.yaml, adjusting Ray version as desired:

    - pip install -U 'ray[default,rllib]==2.0.0'

To run inside docker, see here: Cluster YAML Configuration Options — Ray 2.2.0

cc: @gramhagen , the maintainer for ray on azure