Ray Core x Windows support

Ray Core
1

Hi there,
I am considering using Ray as the parallel compute backend for an open source library in development. Ideally users will be able to easily scale from one core on a single computer to many workers on a distributed cluster. Ray seems like an exciting option for this.

However, we need to support Windows. Is there any particular tracking page or roadmap for known bugs or missing features? Are tests being run on Windows platforms? What else should we know about Windows support for Ray core? I understand that Windows support is in beta, and I’ve browsed through the Github issues page for ray, but I haven’t found any particular organized source of this information or an overview for recent ray releases.

Apologies if I have missed something—first-time ray user. Sorry to ask this pretty broad question, but it’s necessary information for us to make a go/no-go decision at this time. Thanks in advance.

2

The coverage for Windows is fairly robust for the single-node use case. We have a couple folks burning through the Windows issues as we speak. That being said Ray doesn’t work on Windows clusters right now and may have limited support on WSL.

Let me know if that makes sense.

3

Makes sense and is great to hear. Thanks for the quick response!

4

@rliaw

Is the team also working on supporting ray train on windows?

I’m trying to use by spark cluster and convert it to a ray cluster using the example provided here – Deploying on Spark Standalone cluster — Ray 2.6.1

But, I’m getting an error saying only spark on linux is supported currently