[autoscaler] Improve autoscaler auto-configuration, upstream recent improvements to Kuberay NodeProvider

README.md

@@ -26,14 +26,14 @@ You can view detailed documentation and guides at [https://ray-project.github.io
 #### Nightly version
 
 ```
-kubectl apply -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources"
+kubectl create -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources"
 kubectl apply -k "github.com/ray-project/kuberay/manifests/base"
 ```
 
 #### Stable version
 
 ```
-kubectl apply -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources?ref=v0.2.0"
+kubectl create -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources?ref=v0.2.0"
 kubectl apply -k "github.com/ray-project/kuberay/manifests/base?ref=v0.2.0"
 ```
 

docs/deploy/installation.md

@@ -3,13 +3,13 @@
 #### Nightly version
 
 ```
-kubectl apply -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources"
+kubectl create -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources"
 kubectl apply -k "github.com/ray-project/kuberay/manifests/base"
 ```
 
 #### Stable version
 
 ```
-kubectl apply -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources?ref=v0.2.0"
+kubectl create -k "github.com/ray-project/kuberay/manifests/cluster-scope-resources?ref=v0.2.0"
 kubectl apply -k "github.com/ray-project/kuberay/manifests/base?ref=v0.2.0"
 ```

docs/guidance/autoscaler.md

@@ -10,10 +10,14 @@ You can follow below steps for a quick deployment.
 ```
 git clone https://github.com/ray-project/kuberay.git
 cd kuberay
-kubectl apply -k manifests/cluster-scope-resources
-kubectl apply -k manifests/base
+kubectl create -k manifests/cluster-scope-resources
+kubectl apply -k manifests/overlays/autoscaling
 ```
 
+> Note: For compatibility with the Ray autoscaler, the KubeRay Operator's entrypoint
+> must include the flag `--prioritize-workers-to-delete`. The kustomization overlay
+> `manifests/overlays/autoscaling` provided in the last command above adds the necessary flag.
+
 ### Deploy a cluster with autoscaling enabled
 
 ```
@@ -60,20 +64,21 @@ Demands:
 
 #### Known issues and limitations
 
-1. operator will recognize following setting and automatically inject preconfigured autoscaler container to head pod.
-   The service account, role, role binding needed by autoscaler will be created by operator out-of-box.
+1. The operator will recognize the following setting and automatically inject a preconfigured autoscaler container to the head pod.
+   The service account, role, and role binding needed by the autoscaler will be created by the operator out-of-box.
+   The operator will also configure an empty-dir logging volume for the Ray head pod. The volume will be mounted into the Ray and
+   autoscaler containers; this is necessary to support the event logging introduced in [Ray PR #13434](https://github.com/ray-project/ray/pull/13434).
 
     ```
     spec:
       rayVersion: 'nightly'
       enableInTreeAutoscaling: true
     ```
 
-2. head and work images are `rayproject/ray:413fe0`. This image was built based on [commit](https://github.com/ray-project/ray/commit/413fe08f8744d50b439717564709bc0af2f778f1) from master branch. 
-The reason we need to use a nightly version is because autoscaler needs to connect to Ray cluster. Due to ray [version requirements](https://docs.ray.io/en/latest/cluster/ray-client.html#versioning-requirements).
-We determine to use nightly version to make sure integration is working.
+2. The autoscaler image is `rayproject/ray:448f52` which reflects the latest changes from [Ray PR #24718](https://github.com/ray-project/ray/pull/24718/files) in the master branch.
 
-3. Autoscaler image is `kuberay/autoscaler:nightly` which is built from [commit](https://github.com/ray-project/ray/pull/22689/files).
+3. Autoscaling functionality is supported only with Ray versions at least as new as 1.11.0. The autoscaler image used
+is compatible with all Ray versions >= 1.11.0.
 
 ### Test autoscaling
 

docs/notebook/kuberay-on-kind.ipynb

@@ -140,7 +140,7 @@
     }
    ],
    "source": [
-    "!kubectl apply -k \"github.com/ray-project/kuberay/manifests/cluster-scope-resources\"\n",
+    "!kubectl create -k \"github.com/ray-project/kuberay/manifests/cluster-scope-resources\"\n",
     "!kubectl apply -k \"github.com/ray-project/kuberay/manifests/base\""
    ]
   },

manifests/base/kustomization.yaml

@@ -11,7 +11,7 @@ resources:
 images:
 - name: kuberay/apiserver
   newName: kuberay/apiserver
-  newTag: nightly  
+  newTag: nightly
 - name: kuberay/operator
   newName: kuberay/operator
   newTag: nightly

manifests/overlays/autoscaling/kustomization.yaml

@@ -0,0 +1,14 @@
+# This overlay patches in KubeRay operator configuration
+# necessary for Ray Autoscaler support.
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+bases:
+- ../../base
+patches:
+- path: prioritize_workers_to_delete_patch.json
+  target:
+    group: apps
+    version: v1
+    kind: Deployment
+    name: kuberay-operator

manifests/overlays/autoscaling/prioritize_workers_to_delete_patch.json

@@ -0,0 +1,5 @@
+[{
+  "op":"replace",
+  "path":"/spec/template/spec/containers/0/args",
+  "value": ["--prioritize-workers-to-delete"]
+}]

ray-operator/config/samples/ray-cluster.autoscaler.yaml

@@ -8,7 +8,8 @@ metadata:
     # An unique identifier for the head node and workers of this cluster.
   name: raycluster-autoscaler
 spec:
-  rayVersion: 'nightly'
+  rayVersion: '1.12.1'
+  # Ray autoscaler integration is supported only for Ray versions >= 1.11.0
   enableInTreeAutoscaling: true
   ######################headGroupSpecs#################################
   # head group template and specs, (perhaps 'group' is not needed in the name)
@@ -20,7 +21,7 @@ spec:
     # logical group name, for this called head-group, also can be functional
     # pod type head or worker
     # rayNodeType: head # Not needed since it is under the headgroup
-    # the following params are used to complete the ray start: ray start --head --block --redis-port=6379 ...
+    # the following params are used to complete the ray start: ray start --head --block --port=6379 ...
     rayStartParams:
       # Flag "no-monitor" must be set when running the autoscaler in
       # a sidecar container.
@@ -29,17 +30,17 @@ spec:
       node-ip-address: $MY_POD_IP # auto-completed as the head pod IP
       block: 'true'
       num-cpus: '1' # can be auto-completed from the limits
-      redis-password: 'LetMeInRay' # Deprecated since Ray 1.11 due to GCS bootstrapping enabled
       # Use `resources` to optionally specify custom resource annotations for the Ray node.
       # The value of `resources` is a string-integer mapping.
-      # Currently, `resources` must be provided in the unfortunate format demonstrated below.
+      # Currently, `resources` must be provided in the unfortunate format demonstrated below:
+      # resources: '"{\"Custom1\": 1, \"Custom2\": 5}"'
     #pod template
     template:
       spec:
         containers:
         # The Ray head pod
         - name: ray-head
-          image: rayproject/ray:413fe0
+          image: rayproject/ray:1.12.1
           imagePullPolicy: Always
           env:
           - name: CPU_REQUEST
@@ -124,7 +125,7 @@ spec:
           command: ['sh', '-c', "until nslookup $RAY_IP.$(cat /var/run/secrets/kubernetes.io/serviceaccount/namespace).svc.cluster.local; do echo waiting for myservice; sleep 2; done"]
         containers:
         - name: machine-learning # must consist of lower case alphanumeric characters or '-', and must start and end with an alphanumeric character (e.g. 'my-name',  or '123-abc'
-          image: rayproject/ray:413fe0
+          image: rayproject/ray:1.12.1
           # environment variables to set in the container.Optional.
           # Refer to https://kubernetes.io/docs/tasks/inject-data-application/define-environment-variable-container/
           env: