Ray 1.10: Windows support beta, enhanced job submission, and more

By Chandler Gibbons   

Ray version 1.10 is here!

There’s plenty of goodness in this release, but here are the highlights:

  • Windows support, now in beta

  • Enhancements to Ray job submission, including log streaming and custom headers for authentication

You can run pip install -U ray to access these features and more. With that, let’s dive in.

LinkWindows support is now in beta

We’re excited to announce that, with version 1.10, Windows support for Ray is now in beta. Since we first introduced Windows support in alpha in version 0.8.6, we’ve continued to support and maintain the Windows build to make sure more developers can use Ray. Now, a significant fraction of the Ray test suite is passing on Windows.

We want to hear about your experience with Ray 1.10 on Windows! Please file any issues you encounter at https://github.com/ray-project/ray/issues. In upcoming releases, we will continue to invest in making Ray Serve and Runtime Environment tests pass on Windows.

LinkRay job submission enhancements

The goal of Ray job submission is to provide a lightweight mechanism for you to submit locally developed and tested applications to a running remote Ray cluster. You can then package, deploy, and manage your Ray applications as jobs, which can be submitted by a job manager of your choice. Check out our documentation for more on Ray job submission.

With version 1.10, we’re introducing several enhancements to Ray job submission, including jobs log streaming and custom headers and cookies in the job submission client.

First, we’ve added log streaming for a better user experience and to better integrate with external job managers. Previously, there was only one API for fetching job logs, which returned the entire log as a string.

Now, in 1.10, the jobs submission CLI will automatically follow and stream logs for active jobs, and Python and REST APIs return new log lines in real time. This facilitates the custom use of job managers such as Apache Airflow or running jobs on Kubernetes pods, where logs will automatically populate without additional polling and processing from the user side. To get started streaming logs from the job submission server, just use the ray job logs command.

Finally, we’ve also added the ability to propagate a custom headers field to the JobSubmissionClient and apply it to all requests, allowing authentication with a remote Ray cluster. This makes it possible to leverage the job submission feature in data centers or servers where authentication is required.

LinkLearn more

To learn about all the features and enhancements in this release, including continuing progress in improving Ray stability and usability on Windows, check out the Ray 1.10.0 release notes. If you would like to keep up to date with all things Ray, follow @raydistributed on Twitter and sign up for the Ray newsletter.

Ready to try Anyscale?

Access Anyscale today to see how companies using Anyscale and Ray benefit from rapid time-to-market and faster iterations across the entire AI lifecycle.