Helm chart CD and multiple images for worker types

.github/workflows/release.yml

@@ -0,0 +1,26 @@
+name: release
+
+on:
+  release:
+    types: [prereleased, released]
+
+jobs:
+  upload-helm-charts:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v2
+
+      - uses: azure/setup-helm@v1
+
+      - name: Package helm charts
+        run: for dir in ./*/; do helm package $dir; done
+        working-directory: deploy/charts/
+
+      - name: Rename helm chart archives
+        run: for filename in *.tgz; do mv "$filename" "helm-chart-$filename"; done
+        working-directory: deploy/charts/
+
+      - uses: softprops/action-gh-release@v1
+        with:
+          files: deploy/charts/*.tgz

deploy/charts/ray/Chart.yaml

@@ -4,7 +4,7 @@ description: A Helm chart for deployments of Ray on Kubernetes.
 type: application
 
 # Chart version.
-version: 0.1.0
+version: 0.2.0
 
 # Ray version.
 appVersion: "latest"

deploy/charts/ray/templates/raycluster.yaml

@@ -42,7 +42,7 @@ spec:
           containers:
           - name: ray-node
             imagePullPolicy: Always
-            image: {{ $.Values.image }}
+            image: {{ $val.image | default $.Values.image }}
             # Do not change this command - it keeps the pod alive until it is
             # explicitly killed.
             command: ["/bin/bash", "-c", "--"]

deploy/charts/ray/values.yaml

@@ -2,7 +2,8 @@
 
 # RayCluster settings:
 
-# image is Ray image to use for the head and workers of this Ray cluster.
+# image is the default Ray image to use for the head and workers of this Ray cluster
+# You can override image for specific podTypes, e.g. for GPU workers
 # It's recommended to build custom dependencies for your workload into this image,
 # taking one of the offical `rayproject/ray` images as base.
 image: rayproject/ray:latest
@@ -14,12 +15,13 @@ podTypes:
     # Since we set headPodType: rayHeadType, the Ray head pod will use the configuration
     # defined in this entry of podTypes:
     rayHeadType:
+        image: ""
         # CPU is the number of CPUs used by this pod type.
         # (Used for both requests and limits. Must be an integer, as Ray does not support fractional CPUs.)
         CPU: 1
         # memory is the memory used by this Pod type.
         # (Used for both requests and limits.)
-        memory: 512Mi
+        memory: 1Gi
         # GPU is the number of NVIDIA GPUs used by this pod type.
         # (Optional, requires GPU nodes with appropriate setup. See https://docs.ray.io/en/master/cluster/kubernetes-gpu.html)
         GPU: 0
@@ -43,13 +45,14 @@ podTypes:
 
     # The key for each podType is a user-defined string.
     rayWorkerType:
+        image: ""
         # minWorkers is the minimum number of Ray workers of this pod type to keep running.
         minWorkers: 2
         # maxWorkers is the maximum number of Ray workers of this pod type to which Ray will scale.
         maxWorkers: 3
         # memory is the memory used by this Pod type.
         # (Used for both requests and limits.)
-        memory: 512Mi
+        memory: 1Gi
         # CPU is the number of CPUs used by this pod type.
         # (Used for both requests and limits. Must be an integer, as Ray does not support fractional CPUs.)
         CPU: 1

doc/source/cluster/kubernetes.rst

@@ -50,14 +50,15 @@ Installing the Ray Operator with Helm
 Ray provides a `Helm`_ chart to simplify deployment of the Ray Operator and Ray clusters.
 
 The `Ray Helm chart`_ is available as part of the Ray GitHub repository.
+A URL to the packaged chart can be found under the `release assets`_.
 The chart will be published to a public Helm repository as part of a future Ray release.
 
 Preparation
 ~~~~~~~~~~~
 
 - Configure `kubectl`_ to access your Kubernetes cluster.
 - Install `Helm 3`_.
-- Download the `Ray Helm chart`_.
+- Download the `Ray Helm chart`_, or find the URL to the packaged Helm chart under `release assets`_.
 
 To run the default example in this document, make sure your Kubernetes cluster can accomodate
 additional resource requests of 4 CPU and 2.5Gi memory.
@@ -66,6 +67,8 @@ Installation
 ~~~~~~~~~~~~
 
 You can install a small Ray cluster with a single ``helm`` command.
+You can either clone the project and point Helm to the local path,
+or point Helm to the packaged chart by passing the URL found under `release assets`_.
 The default cluster configuration consists of a Ray head pod and two worker pods,
 with scaling allowed up to three workers.
 
@@ -306,6 +309,7 @@ Questions or Issues?
 .. _`namespace`: https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/
 .. _`Deployment`: https://kubernetes.io/docs/concepts/workloads/controllers/deployment/
 .. _`Ray Helm chart`: https://github.com/ray-project/ray/tree/master/deploy/charts/ray/
+.. _`release assets`: https://github.com/ray-project/ray/releases/latest/
 .. _`kubectl`: https://kubernetes.io/docs/tasks/tools/
 .. _`Helm 3`: https://helm.sh/
 .. _`Helm`: https://helm.sh/