Merge branch 'main' into geniac

Author: KeitaW

0.docs/fsx-lustre-template.png

@@ -42,11 +42,11 @@ The templates takes parameters that are mandatory and optional, see below for mo
 If you'd like to deploy through the AWS CLI instead of the quick create link above, the command to deploy the template is shown below. Please edit the parameters values with your own configuration.
 
 ```bash
-aws cloudformation create-stack --stack-name batch-distributed-training \
-                                --template-body file://0.aws-batch-distributed-training.yaml \
-                                --parameters ParameterKey=VPCStackParameter,ParameterValue="vpc-stack-ml" \
-                                             ParameterKey=CapacityReservationId,ParameterValue="cr-123567890abc" \
-                                --capabilities CAPABILITY_IAM
+aws cloudformation create-stack --stack-name aws-batch-p5 \
+                                --template-body file://0.aws-batch-distributed-training-p5.yaml \
+                                --parameters ParameterKey=VPCStackParameter,ParameterValue="aws-batch-vpc" \
+                                             ParameterKey=CapacityReservationId,ParameterValue="cr-1234567890" \
+                                --capabilities CAPABILITY_NAMED_IAM
 ```
 
 ## Gotchas

1.architectures/3.aws-batch/0.aws-batch-distributed-training-p5.yaml

@@ -171,7 +171,7 @@ kubectl get nodes
 5. Apply [K8 Nvidia CNI Plugin](https://github.com/NVIDIA/k8s-device-plugin):
 
 ```bash
-kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.15.0/deployments/static/nvidia-device-plugin.yml
+kubectl create -f https://raw.githubusercontent.com/NVIDIA/k8s-device-plugin/v0.17.0/deployments/static/nvidia-device-plugin.yml
 ```
 
 6. If using EFA, make sure to install the [EFA CNI Plugin](https://docs.aws.amazon.com/eks/latest/userguide/node-efa.html).

1.architectures/3.aws-batch/README.md

@@ -1,6 +1,6 @@
 AWSTemplateFormatVersion: "2010-09-09"
 
-Description: Amazon EKS - Create an unmanaged P4d/P5 node group for Capacity Blocks for ML.
+Description: Amazon EKS - Create an unmanaged P4d/P5/P5e node group for Capacity Blocks for ML.
 
 Metadata:
   "AWS::CloudFormation::Interface":
@@ -75,7 +75,7 @@ Parameters:
 
   NodeImageIdSSMParam:
     Type: "AWS::SSM::Parameter::Value<AWS::EC2::Image::Id>"
-    Default: /aws/service/eks/optimized-ami/1.29/amazon-linux-2-gpu/recommended/image_id
+    Default: /aws/service/eks/optimized-ami/1.31/amazon-linux-2-gpu/recommended/image_id
     Description: AWS Systems Manager Parameter Store parameter of the AMI ID for the worker node instances. Change this value to match the version of Kubernetes you are using.
 
   DisableIMDSv1:
@@ -89,6 +89,7 @@ Parameters:
     Type: String
     Default: p5.48xlarge
     AllowedValues:
+      - p5e.48xlarge
       - p5.48xlarge
       - p4d.24xlarge
     Description: EC2 instance type for the node instances
@@ -134,12 +135,16 @@ Conditions:
     - "Fn::Equals":
       - !Ref NodeImageId
       - ""
-  isP5: !Equals
-      - !Ref NodeInstanceType
-      - "p5.48xlarge"
   isP4d: !Equals
       - !Ref NodeInstanceType
       - "p4d.24xlarge"
+  isP5Family: !Or
+    - !Equals
+      - !Ref NodeInstanceType
+      - "p5.48xlarge"
+    - !Equals
+      - !Ref NodeInstanceType
+      - "p5e.48xlarge"
 
   IMDSv1Disabled:
     "Fn::Equals":
@@ -280,7 +285,7 @@ Resources:
 
   NodeLaunchTemplateP5:
     Type: "AWS::EC2::LaunchTemplate"
-    Condition: isP5
+    Condition: isP5Family
     Properties:
       LaunchTemplateData:
         InstanceMarketOptions:
@@ -659,11 +664,11 @@ Resources:
       DesiredCapacity: !Ref NodeAutoScalingGroupDesiredCapacity
       LaunchTemplate:
         LaunchTemplateId: !If
-          - isP5
+          - isP5Family
           - !Ref NodeLaunchTemplateP5
           - !Ref NodeLaunchTemplateP4
         Version: !If
-          - isP5
+          - isP5Family
           - !GetAtt NodeLaunchTemplateP5.LatestVersionNumber
           - !GetAtt NodeLaunchTemplateP4.LatestVersionNumber
       MaxSize: !Ref NodeAutoScalingGroupMaxSize

1.architectures/4.amazon-eks/README.md

@@ -1,7 +1,6 @@
 AWSTemplateFormatVersion: '2010-09-09'
-Description: Creates an FSxL filesystem of PERSISENT_2 type plus the Security Group needed for use with SageMaker
+Description: This template deploys an FSx for Luster File System
 
-### Stack metadata
 Metadata:
   AWS::CloudFormation::Interface:
     ParameterGroups:
@@ -13,15 +12,18 @@ Metadata:
           - Compression
           - LustreVersion
 
+      - Label:
+          default: Networking Options
+        Parameters:
+          - SecurityGroup
+          - Subnet
+
 Parameters:
-  NetworkStack:
-    Description: Name of the Networking stack
-    Type: String
-    Default: SageMakerVPC
   Capacity:
     Description: Storage capacity in GiB (1200 or increments of 2400)
     Type: Number
     Default: 1200
+
   PerUnitStorageThroughput:
     Description: Provisioned Read/Write (MB/s/TiB)
     Type: Number
@@ -31,13 +33,15 @@ Parameters:
       - 250
       - 500
       - 1000
+
   Compression:
     Description: Data compression type
     Type: String
     AllowedValues:
       - "LZ4"
       - "NONE"
     Default: "LZ4"
+
   LustreVersion:
     Description: Lustre software version
     Type: String
@@ -46,27 +50,18 @@ Parameters:
       - "2.12"
     Default: "2.15"
 
-Resources:
+  SecurityGroup:
+    Description: Security group ID
+    Type: String
+    Default: ""
+
+  Subnet:
+    Description: Subnet ID
+    Type: String
+    Default: ""
 
-  LambdaExecutionRole:
-    Type: "AWS::IAM::Role"
-    Properties:
-      AssumeRolePolicyDocument:
-        Version: 2012-10-17
-        Statement:
-          - Effect: Allow
-            Principal:
-              Service:
-                - lambda.amazonaws.com
-            Action:
-              - "sts:AssumeRole"
-      Path: /
-      ManagedPolicyArns:
-        - 'arn:aws:iam::aws:policy/AmazonEC2ReadOnlyAccess'
-        - 'arn:aws:iam::aws:policy/AmazonS3FullAccess'
-        - 'arn:aws:iam::aws:policy/IAMFullAccess'
-        - 'arn:aws:iam::aws:policy/service-role/AWSLambdaBasicExecutionRole'
 
+Resources:
   FSxLFilesystem:
     Type: AWS::FSx::FileSystem
     DeletionPolicy: Delete
@@ -77,22 +72,29 @@ Resources:
       FileSystemTypeVersion: !Ref LustreVersion
       StorageCapacity: !Ref Capacity
       SecurityGroupIds:
-        - Fn::ImportValue:
-            !Sub "${NetworkStack}-SecurityGroup"
+        - !Ref SecurityGroup
       SubnetIds:
-        - Fn::ImportValue:
-            !Sub "${NetworkStack}-PrivateSubnet"
+        - !Ref Subnet
       LustreConfiguration:
         DataCompressionType: !Ref Compression
         DeploymentType: PERSISTENT_2
         PerUnitStorageThroughput: !Ref PerUnitStorageThroughput
+        MetadataConfiguration:
+          Mode: AUTOMATIC
 
 Outputs:
   FSxLustreFilesystemMountname:
     Description: The ID of the FSxL filesystem that has been created
     Value: !GetAtt FSxLFilesystem.LustreMountName
     Export:
       Name: !Sub ${AWS::StackName}-FSxLustreFilesystemMountname
+
+  FSxLustreFilesystemDNSname:
+    Description: The DNS of the FSxL filesystem that has been created
+    Value: !GetAtt FSxLFilesystem.DNSName
+    Export:
+      Name: !Sub ${AWS::StackName}-FSxLustreFilesystemDNSname
+
   FSxLustreFilesystemId:
     Description: The ID of the FSxL filesystem that has been created
     Value: !Ref FSxLFilesystem

1.architectures/4.amazon-eks/amazon-eks-nodegroup.yaml

@@ -2,7 +2,7 @@
 # Basic configuration parameters
 class Config:
 
-    # Set true if you want to install Docker/Enroot/Pyxis.
+    # Default is true to install Docker/Enroot/Pyxis.
     enable_docker_enroot_pyxis = True
 
     # Set true if you want to install metric exporter software and Prometheus for observability
@@ -24,8 +24,16 @@ class Config:
     # You need to configure parameters in SssdConfig as well.
     enable_sssd = False
 
-    # Set true to install quality-of-live improvements
-    enable_initsmhp = False
+    # Set true if you want to use mountpoint for s3 on cluster nodes. 
+    # If enabled, a systemctl mount-s3.service file will be writen that will mount at /mnt/<BucketName>.
+    # requires s3 permissions to be added to cluster execution role. 
+    enable_mount_s3 = False
+
+    s3_bucket = "" # required when enable_mount_s3 = True, replace with your actual data bucket name in quotes, ie. "my-dataset-bucket"
+
+    if enable_mount_s3 and not s3_bucket:
+        raise ValueError("Error: A bucket name must be specified when enable_mount_s3 is True")
+
 
 # Configuration parameters for ActiveDirectory/LDAP/SSSD
 class SssdConfig: