Hey, I have a specific use case where I would like to redeploy ray serve applications due to changes in model config that can’t be modified through a reconfigure() method. Is there a simple way to force a redeploy of worker nodes on K8’s.
Hello rkunds! Welcome to the Ray community~ 
If you need to redeploy your Ray Serve applications on K8, there’s a few ways to do it if you’re avoiding reconfigure(). You can trigger a redeployment by modifying certain deployment configuration values.
For instance, changing parameters like ray_actor_options , placement_group_bundles , placement_group_strategy , import_path , or runtime_env will cause the system to redeploy and restart all the replicas of your deployment.
However, if you’re planning a big update, such as a large-scale code change, the best practice is to set up a new Ray cluster. You can deploy your updated code to this new cluster using serve deploy , and then transition traffic from your old cluster to the new one.
Here’s a few docs that might help out. Let me know if these were helpful.
Docs: