update cluster resources

CHANGELOG.md

@@ -19,6 +19,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### **Changed**
 - changed `ray-image` to pull from AWS Public ECR to avoid docker pull rate limits
 - changed `ray-orchestrator` to not retrieve full training job logs and avoid `States.DataLimitExceeded`
+- update `ray-on-eks` manifest cluster resources
 
 ## v1.7.0
 

manifests/ray-on-eks/core-modules.yaml

@@ -44,9 +44,9 @@ parameters:
           eks_node_labels:
             usage: core
         - eks_ng_name: ng-gpu
-          eks_node_quantity: 1
+          eks_node_quantity: 5
           eks_node_max_quantity: 10
-          eks_node_min_quantity: 1
+          eks_node_min_quantity: 5
           eks_node_disk_size: 400
           eks_node_instance_type: "g4dn.4xlarge"
           eks_node_labels:

manifests/ray-on-eks/ray-cluster-modules.yaml

@@ -31,11 +31,11 @@ parameters:
   - name: HeadResources
     value:
       requests:
-        cpu: "1"
-        memory: "8G"
+        cpu: "8"
+        memory: "24G"
       limits:
-        cpu: "4"
-        memory: "16G"
+        cpu: "8"
+        memory: "24G"
   - name: WorkerReplicas
     value: 1
   - name: WorkerMinReplicas
@@ -45,8 +45,8 @@ parameters:
   - name: WorkerResources
     value:
       requests:
-        cpu: "4"
-        memory: "8G"
+        cpu: "14"
+        memory: "60G"
       limits:
         cpu: "14"
         memory: "60G"