aws-samples / appstream-data-tools-isolation-blog Goto Github PK

We had deployed the provided solution in our environment and observed an issue with UserPool user.

While executing the solution with custom user we could see Appstream describe session is able to generate active session Id in lambda function and able to start the sagemaker notebook successfully.

Observed an issue while executing with UserPool user as the lambda function describe-session is not able to generate active sessions and leads lambda to failure with below error

[ERROR] IndexError: list index out of range Traceback (most recent call last):
File "/var/task/data_sandbox_lambda.py", line 27, in lambda_handler
resp_user_session = resp['Sessions'][0]['Id']

Attached error files
Cloudwatch log shows both custom & userpool user

Active session logs from CLI.

Thanks,
Pavan

With CDK V2 becoming GA, it became the default CDK version installed when running the command npm install -g aws-cdk.

I suggest creating a package.json file that tells npm to locally install CDK with a specific version.

{
    "devDependencies": {
        "aws-cdk": "1.X.0"
    }
}

The commands used to run this solution would changed like so:

Change npm install -g aws-cdk to npm install.
Change cdk bootstrap to npx cdk bootstrap.
Change cdk deploy DataSandbox to npx cdk deploy DataSandbox.

I tried to deploy this solution and I encountered a problem that caused the data_sandbox_lambda Lambda function to fail at put_object with error:

[ERROR] NoSuchBucket: An error occurred (NoSuchBucket) when calling the PutObject operation: The specified bucket does not exist

How to reproduce:

At the image creation step, create an AppStream 2.0 image with digits in the name, for example my-image-00. Then continue as normal through the steps of deployment as described in the blog post.

Deeper dive into the problem:

I debugged it for a bit and found that in the session.json file, "bucketName" was:

appstream2-36fb080bb8-us-east-1-12345678901200

Where my account id would be 123456789012 (notice the trailing "00").

I searched for where these trailing zeros come from and found that in the sagemaker-notebook.ps1 powershell script, we get the account ID like so:

$ArnID = $env:AppStream_Image_Arn -replace "[^0-9]", ''
$AccountId = $ArnID.substring(1)

Which does not take into account digits in the image name, and would leave trailing digits in the account id if any are present.

Resolution

To overcome this problem I changed the two lines above to with this line:

$AccountId = $env:AppStream_Image_Arn.Split(":")[4]

Description

When install the python dependencies using

pip install -r requirements.txt

I get the following error:

ERROR: Cannot install -r requirements.txt (line 11), -r requirements.txt (line 7) and aws-cdk.aws-s3-assets==1.51.0 because these package versions have conflicting dependencies.

The conflict is caused by:
    The user requested aws-cdk.aws-s3-assets==1.51.0
    aws-cdk-aws-lambda 1.51.0 depends on aws-cdk.aws-s3-assets==1.51.0
    aws-cdk-aws-s3-deployment 1.50.0 depends on aws-cdk.aws-s3-assets==1.50.0

There is a conflict between aws-cdk-aws-lambda 1.51.0 and aws-cdk-aws-s3-deployment 1.50.0.

Upgrading aws-cdk-aws-s3-deployment 1.50.0 to aws-cdk-aws-s3-deployment 1.51.0 works, however after doing that, when deploying the CDK application I get to following error:

Embedded stack arn:aws:cloudformation:us-east-1:012345678901:stack/DataSandbox-appstreamservicerolesstackNestedStackappstreamservicerolesstackNestedStack-9D140RXJRAJN/2a1e76b0-c3b1-11ec-b6a2-12b935dbbaef was not successfully created: The following resource(s) failed to create: [LogRetentionaae0aa3c5b4d4f87b02d85b201efdd8aFD4BFC8A]. 

Resolution

To overcome this I upgraded all the aws-cdk* python packages to the current latest CDK V1 version: 1.153.1.