cdk-efs-assets

CDK construct library to populate Amazon EFS assets from Github or S3. If the source is S3, the construct also optionally supports updating the contents in EFS if a new zip file is uploaded to S3.

Install

TypeScript/JavaScript:

1npm i cdk-efs-assets

SyncedAccessPoint

The main construct that is used to provide this EFS sync functionality is SyncedAccessPoint. This extends the standard EFS AccessPoint construct, and takes an additional SyncSource constructor property which defines the source to sync assets from. The SyncedAccessPoint instance can be used anywhere an AccessPoint can be used. For example, to specify a volume in a Task Definition:

1const taskDefinition = new ecs.FargateTaskDefinition(this, 'TaskDefinition', {
2  ... 
3  volumes: [
4    {
5      name: 'efs-storage',
6      efsVolumeConfiguration: {
7        fileSystemId: sharedFileSystem.fileSystemId,
8        transitEncryption: 'ENABLED',
9        authorizationConfig: {
10          accessPointId: syncedAccessPoint.accessPointId
11        }
12      }
13    },
14  ]
15});

Sync Engine

This library supports both AWS Fargate and AWS Lambda as the sync engine. As AWS Lambda currently has know issue with Amazon EFS(#100), the default sync engine is AWS Fargate. You can opt in AWS Lambda with the engine construct property of SyncedAccessPoint.

Sync Source

Use GithubSyncSource and S3ArchiveSyncSource construct classes to define your syncSource from Github or Amazon S3 bucket. For example:

To define a public github repository as the syncSource:

1new SyncedAccessPoint(stack, 'EfsAccessPoint', {
2  ...
3  syncSource: new GithubSyncSource({
4    vpc,
5    repository: 'https://github.com/pahud/cdk-efs-assets.git',
6  }),
7});

To define a private github repository as the syncSource:

1new SyncedAccessPoint(stack, 'EfsAccessPoint', {
2  ...
3  syncSource: new GithubSyncSource({
4    vpc,
5    repository: 'https://github.com/pahud/private-repo.git',
6    secret: {
7      id: 'github',
8      key: 'oauth_token',
9    },
10  }),
11});

syncDirectoryPath

By default, the synced EFS assets are placed into a directory corresponding to the type of the sync source. For example, the default behavior of the GitHub source is to place the copied files into a directory named the same as the repository name (for a repository specified as 'https://github.com/pahud/cdk-efs-assets.git', the directory name would be 'cdk-efs-assets'), while the default behavior of the S3 archive source is to place the copied files into a directory named the same as the zip file (for a zip file name of 'assets.zip', the directory name would be 'assets').

If you wish to override this default behavior, specify a value for the syncDirectoryPath property that is passed into the SyncSource call.

If you are using the AccessPoint in an ECS/Fargate Task Definition, you probably will want to override the value of syncDirectoryPath to '/'. This will place the file contents in the root directory of the Access Point. The reason for this is that when you create a volume that is referencing an EFS Access Point, you are not allowed to specify any path other than the root directory in the task definition configuration.

How to use SyncedAccessPoint initialized with files provisioned from GitHub repository

This will sync assets from a GitHub repository to a directory (by default, the output directory is named after the repository name) in the EFS AccessPoint:

1import { SyncSource, SyncedAccessPoint } from 'cdk-efs-assets';
2
3const app = new App();
4
5const env = {
6  region: process.env.CDK_DEFAULT_REGION ?? AWS_DEFAULT_REGION,
7  account: process.env.CDK_DEFAULT_ACCOUNT,
8};
9
10const stack = new Stack(app, 'testing-stack', { env });
11
12const vpc = ec2.Vpc.fromLookup(stack, 'Vpc', { isDefault: true })
13
14const fileSystem = new efs.FileSystem(stack, 'Filesystem', {
15  vpc,
16  removalPolicy: RemovalPolicy.DESTROY,
17})
18
19const efsAccessPoint = new SyncedAccessPoint(stack, 'GithubAccessPoint', {
20  vpc,
21  fileSystem,
22  path: '/demo-github',
23  createAcl: {
24    ownerGid: '1001',
25    ownerUid: '1001',
26    permissions: '0755',
27  },
28  posixUser: {
29    uid: '1001',
30    gid: '1001',
31  },
32  syncSource: new GithubSyncSource({
33    vpc,
34    repository: 'https://github.com/pahud/cdk-efs-assets.git',
35  })
36});

Github private repository support

To clone a github private repository, you need to generate your github PAT(Personal Access Token) and store the token in AWS Secrets Manager secret.

For example, if your PAT is stored in the AWS Secret manager with the secret ID github and the key oauth_token, you can specify the secret property as the sample below. Under the covers the lambda function will format the full github repository uri with your PAT and successfully git clone the private repository to the efs filesystem.

Store your PAT into the AWS Secrets Manager with AWS CLI:

1aws secretsmanager create-secret \
2--name github \
3--secret-string '{"oauth_token":"MYOAUTHTOKEN"}'

Configure the secret property to allow lambda to retrieve the PAT from the secret:

1new GithubSyncSource({
2    vpc,
3    repository: 'https://github.com/username/repo.git',
4    secret: {
5      id: 'github',
6      key: 'oauth_token',
7    },
8})

How to use SyncedAccessPoint initialized with files provisioned from zip file stored in S3

This will sync assets from a zip file stored in an S3 bucket to a directory (by default, the output directory is named after the zip file name) in the EFS AccessPoint:

1import { S3ArchiveSync } from 'cdk-efs-assets';
2
3const app = new App();
4
5const env = {
6  region: process.env.CDK_DEFAULT_REGION ?? AWS_DEFAULT_REGION,
7  account: process.env.CDK_DEFAULT_ACCOUNT,
8};
9
10const stack = new Stack(app, 'testing-stack', { env });
11
12const vpc = ec2.Vpc.fromLookup(stack, 'Vpc', { isDefault: true })
13
14const fileSystem = new efs.FileSystem(stack, 'Filesystem', {
15  vpc,
16  removalPolicy: RemovalPolicy.DESTROY,
17})
18
19const bucket = Bucket.fromBucketName(this, 'Bucket', 'demo-bucket');
20
21const efsAccessPoint = new SyncedAccessPoint(stack, 'EfsAccessPoint', {
22  vpc,
23  fileSystem,
24  path: '/demo-s3',
25  createAcl: {
26    ownerGid: '1001',
27    ownerUid: '1001',
28    permissions: '0755',
29  },
30  posixUser: {
31    uid: '1001',
32    gid: '1001',
33  },
34  syncSource: new S3ArchiveSyncSource({
35    vpc,
36    bucket,
37    zipFilePath: 'folder/foo.zip',
38  }),
39});

syncOnUpdate

If the syncOnUpdate property is set to true (defaults to true), then the specified zip file path will be monitored, and if a new object is uploaded to the path, then it will resync the data to EFS. Note that to use this functionality, you must have a CloudTrail Trail in your account that captures the desired S3 write data event.

This feature is only available with the LAMBDA sync engine.

WARNING: The contents of the extraction directory in the access point will be destroyed before extracting the zip file.

StatefulFargateNginx

This library comes with StatefulFargateNginx construct which allows you to build an Amazon EFS-backed stateful AWS Fargate service with its document root seeded from any github repository.

See this tweet for the demo.

Sample:

1new StatefulFargateNginx(this, 'NyanCat', {
2  vpc,
3  github: 'https://github.com/cristurm/nyan-cat.git',
4});
5