Debug applications and jobs with EMR Studio - Amazon EMR
Amazon EMR running on EC2 jobsAmazon EMR Serverless applicationsAmazon EMR on EKS job runs

Debug applications and jobs with EMR Studio

With Amazon EMR Studio, you can launch data application interfaces to analyze applications and job runs in the browser.

You can also launch the persistent, off-cluster user interfaces for Amazon EMR running on EC2 clusters from the Amazon EMR console. For more information, see View persistent application user interfaces.

Note

Depending on your browser settings, you might need to enable pop-ups for an application UI to open.

For information about configuring and using the application interfaces, see The YARN Timeline Server, Monitoring and instrumentation, or Tez UI overview.

Debug Amazon EMR running on Amazon EC2 jobs

Workspace UI
Launch an on-cluster UI from a notebook file

When you use Amazon EMR release versions 5.33.0 and later, you can launch the Spark web user interface (the Spark UI or Spark History Server) from a notebook in your Workspace.

On-cluster UIs work with the PySpark, Spark, or SparkR kernels. The maximum viewable file size for Spark event logs or container logs is 10 MB. If your log files exceed 10 MB, we recommend that you use the persistent Spark History Server instead of the on-cluster Spark UI to debug jobs.

Important

In order for EMR Studio to launch on-cluster application user interfaces from a Workspace, a cluster must be able to communicate with the Amazon API Gateway. You must configure the EMR cluster to allow outgoing network traffic to Amazon API Gateway, and make sure that Amazon API Gateway is reachable from the cluster.

The Spark UI accesses container logs by resolving hostnames. If you use a custom domain name, you must make sure that the hostnames of your cluster nodes can be resolved by Amazon DNS or by the DNS server you specify. To do so, set the Dynamic Host Configuration Protocol (DHCP) options for the Amazon Virtual Private Cloud (VPC) that is associated with your cluster. For more information about DHCP options, see DHCP option sets in the Amazon Virtual Private Cloud User Guide.

  1. In your EMR Studio, open the Workspace that you want to use and make sure that it is attached to an Amazon EMR cluster running on EC2. For instructions, see Attach a compute to an EMR Studio Workspace.

  2. Open a notebook file and use the PySpark, Spark, or SparkR kernel. To select a kernel, choose the kernel name from the upper right of the notebook toolbar to open the Select Kernel dialog box. The name appears as No Kernel! if no kernel has been selected.

  3. Run your notebook code. The following appears as output in the notebook when you start the Spark context. It might take a few seconds to appear. If you have started the Spark context, you can run the %%info command to access a link to the Spark UI at any time.

    Note

    If the Spark UI links do not work or do not appear after a few seconds, create a new notebook cell and run the %%info command to regenerate the links.

    Screenshot of the Spark application master information, with link to the Spark UI. The link appears in a notebook when you run a Spark application.

  4. To launch the Spark UI, choose Link under Spark UI. If your Spark application is running, the Spark UI opens in a new tab. If the application has completed, the Spark History Server opens instead.

    After you launch the Spark UI, you can modify the URL in the browser to open the YARN ResourceManager or the Yarn Timeline Server. Add one of the following paths after amazonaws.com.

    Web UI Path Example modified URL
    YARN ResourceManager /rm https://j-examplebby5ij.emrappui-prod.eu-west-1.amazonaws.com/rm
    Yarn Timeline Server /yts https://j-examplebby5ij.emrappui-prod.eu-west-1.amazonaws.com/yts
    Spark History Server /shs https://j-examplebby5ij.emrappui-prod.eu-west-1.amazonaws.com/shs
Studio UI
Launch the persistent YARN Timeline Server, Spark History Server, or Tez UI from the EMR Studio UI

  1. In your EMR Studio, select Amazon EMR on EC2 on the left side of the page to open the Amazon EMR on EC2 clusters list.

  2. Filter the list of clusters by name, state, or ID by entering values in the search box. You can also search by creation time range.

  3. Select a cluster and then choose Launch application UIs to select an application user interface. The Application UI opens in a new browser tab and might take some time to load.

Debug EMR Studio running on EMR Serverless

Similar to Amazon EMR running on Amazon EC2, you can use the Workspace user interface to analyze your EMR Serverless applications. From the Workspace UI, when you use Amazon EMR releases 6.14.0 and higher, you can launch the Spark web user interface (the Spark UI or Spark History Server) from a notebook in your Workspace. For your convenience, we also provide a link to the driver log for quick access the Spark driver logs.

Debug Amazon EMR on EKS job runs with the Spark History Server

When you submit a job run to an Amazon EMR on EKS cluster, you can access logs for that job run using the Spark History Server. The Spark History Server provides tools for monitoring Spark applications, such as a list of scheduler stages and tasks, a summary of RDD sizes and memory usage, and environmental information. You can launch the Spark History Server for Amazon EMR on EKS job runs in the following ways:

  • When you submit a job run using EMR Studio with an Amazon EMR on EKS managed endpoint, you can launch the Spark History Server from a notebook file in your Workspace.

  • When you submit a job run using the AWS CLI or AWS SDK for Amazon EMR on EKS, you can launch the Spark History Server from the EMR Studio UI.

For information about how to use the Spark History Server, see Monitoring and Instrumentation in the Apache Spark documentation. For more information about job runs, see Concepts and components in the Amazon EMR on EKS Development Guide.

To launch the Spark History Server from a notebook file in your EMR Studio Workspace

  1. Open a Workspace that is connected to an Amazon EMR on EKS cluster.

  2. Select and open your notebook file in the Workspace.

  3. Choose Spark UI at the top of the notebook file to open the persistent Spark History Server in a new tab.

To launch the Spark History Server from the EMR Studio UI
Note

The Jobs list in the EMR Studio UI displays only job runs that you submit using the AWS CLI or AWS SDK for Amazon EMR on EKS.

  1. In your EMR Studio, select Amazon EMR on EKS on the left side of the page.

  2. Search for the Amazon EMR on EKS virtual cluster that you used to submit your job run. You can filter the list of clusters by status or ID by entering values in the search box.

  3. Select the cluster to open its detail page. The detail page displays information about the cluster, such as ID, namespace, and status. The page also shows a list of all the job runs submitted to that cluster.

  4. From the cluster detail page, select a job run to debug.

  5. In the upper right of the Jobs list, choose Launch Spark History Server to open the application interface in a new browser tab.