| 1 | import * as AWS from "alchemy/AWS"; |
| 2 | import * as Output from "alchemy/Output"; |
| 3 | import * as Effect from "effect/Effect"; |
| 4 | |
| 5 | /** |
| 6 | * Shared infrastructure for the HyperPod cluster. |
| 7 | * |
| 8 | * Resources are memoized by logical id, so this Effect can be yielded from |
| 9 | * the stack program and from any other module and always converges on the |
| 10 | * same single resource instances. |
| 11 | */ |
| 12 | export const HyperPodInfra = Effect.gen(function* () { |
| 13 | /** |
| 14 | * The bucket holding the cluster's lifecycle scripts. Every HyperPod |
| 15 | * instance group downloads and runs its `OnCreate` script from S3 when a |
| 16 | * node boots. |
| 17 | */ |
| 18 | const bucket = yield* AWS.S3.Bucket("LifecycleScripts", { |
| 19 | forceDestroy: true, |
| 20 | }); |
| 21 | |
| 22 | /** |
| 23 | * The instance execution role. HyperPod nodes assume it to pull the |
| 24 | * lifecycle script (inline grant on the bucket) and to publish logs and |
| 25 | * metrics (the AWS-managed cluster-instance policy). |
| 26 | */ |
| 27 | const role = yield* AWS.IAM.Role("HyperPodInstanceRole", { |
| 28 | assumeRolePolicyDocument: { |
| 29 | Version: "2012-10-17", |
| 30 | Statement: [ |
| 31 | { |
| 32 | Effect: "Allow", |
| 33 | Principal: { Service: "sagemaker.amazonaws.com" }, |
| 34 | Action: ["sts:AssumeRole"], |
| 35 | }, |
| 36 | ], |
| 37 | }, |
| 38 | managedPolicyArns: [ |
| 39 | "arn:aws:iam::aws:policy/AmazonSageMakerClusterInstanceRolePolicy", |
| 40 | ], |
| 41 | inlinePolicies: { |
| 42 | "lifecycle-scripts": { |
| 43 | Version: "2012-10-17", |
| 44 | Statement: [ |
| 45 | { |
| 46 | Effect: "Allow", |
| 47 | Action: ["s3:GetObject", "s3:ListBucket"], |
| 48 | Resource: [ |
| 49 | bucket.bucketArn, |
| 50 | Output.interpolate`${bucket.bucketArn}/*`, |
| 51 | ], |
| 52 | }, |
| 53 | ], |
| 54 | }, |
| 55 | }, |
| 56 | }); |
| 57 | |
| 58 | return { bucket, role }; |
| 59 | }); |