Debug lifecycle configurations

The following topics show how to get information about and debug your lifecycle configurations.

Verify lifecycle configuration process from CloudWatch Logs

Lifecycle configurations only log STDOUT and STDERR.

STDOUT is the default output for bash scripts. You can write to STDERR by appending >&2 to the end of a bash command. For example, echo 'hello'>&2.

Logs for your lifecycle configurations are published to your AWS account using Amazon CloudWatch. These logs can be found in the /aws/sagemaker/studio log stream in the CloudWatch console.

1. Open the CloudWatch console at https://console.aws.amazon.com/cloudwatch/.

2. Choose Logs from the left side. From the dropdown menu, select Log groups.

3. On the Log groups page, search for aws/sagemaker/studio.

4. Select the log group.

5. On the Log group details page, choose the Log streams tab.

6. To find the logs for a specific app, search the log streams using the following format:

   domain-id/user-profile-name/app-type/app-name

   For example, to find the lifecycle configuration logs for domain d-m85lcu8vbqmz, user profile i-sonic-js, application type JupyterServer and application name test-lcc-echo, use the following search string:

   d-m85lcu8vbqmz/i-sonic-js/JupyterServer/test-lcc-echo 

7. Select the log stream appended with LifecycleConfigOnStart to view the script execution logs.

JupyterServer app failure

If your JupyterServer app crashes because of an issue with the attached lifecycle configuration, Studio displays the following error message on the Studio startup screen.

Failed to create SageMaker Studio due to start-up script failure

Select the View script logs link to view the CloudWatch logs for your JupyterServer app.

In the case where the faulty lifecycle configuration is specified in the DefaultResourceSpec of your domain, user profile, or shared space, Studio continues to use the lifecycle configuration even after restarting Studio.

To resolve this error, follow the steps in Set default lifecycle configurations to remove the lifecycle configuration script from the DefaultResourceSpec or select another script as the default. Then launch a new JupyterServer app.

KernelGateway app failure

If your KernelGateway app crashes because of an issue with the attached lifecycle configuration, Studio displays the error message in your Studio Notebook.

Choose View script logs to view the CloudWatch logs for your KernelGateway app.

In this case, your lifecycle configuration is specified in the Studio Launcher when launching a new Studio Notebook.

To resolve this error, use the Studio launcher to select a different lifecycle configuration or select No script.

Note

A default KernelGateway lifecycle configuration specified in DefaultResourceSpec applies to all KernelGateway images in the domain, user profile, or shared space unless the user selects a different script from the list presented in the Studio launcher. The default script also runs if No Script is selected by the user. For more information on selecting a script, see Step 3: Launch an application with the lifecycle configuration.

Lifecycle configuration timeout

There is a lifecycle configuration timeout limitation of 5 minutes. If a lifecycle configuration script takes longer than 5 minutes to run, Studio throws an error.

To resolve this error, ensure that your lifecycle configuration script completes in less than 5 minutes.

To help decrease the run time of scripts, try the following:

• Cut down on necessary steps. For example, limit which conda environments to install large packages in.

• Run tasks in parallel processes.

• Use the nohup command in your script to ensure that hangup signals are ignored and do not stop the execution of the script.