# Amazon EMR on EKS 6.8.0 releases
<a name="emr-eks-6.8.0"></a>

The following Amazon EMR 6.8.0 releases are available for Amazon EMR on EKS. Select a specific **emr-6.8.0-XXXX** release to view more details such as the related container image tag.
+ [emr-6.8.0-latest](emr-eks-6.8.0-latest.md)
+  [emr-6.8.0-20230905](emr-eks-6.8.0-20230905.md) 
+ [emr-6.8.0-20230624](emr-eks-6.8.0-20230624.md)
+ [emr-6.8.0-20221219](emr-eks-6.8.0-20221219.md)
+ [emr-6.8.0-20220802](emr-eks-6.8.0-20220802.md)

**Release notes for Amazon EMR 6.8.0**
+ Supported applications ‐ AWS SDK for Java 1.12.170, Spark 3.3.0-amzn-0, Hudi 0.11.1-amzn-0, Iceberg 0.14.0-amzn-0.
+ Supported components ‐ `aws-sagemaker-spark-sdk`, `emr-ddb`, `emr-goodies`, `emr-s3-select`, `emrfs`, `hadoop-client`, `hudi`, `hudi-spark`, `iceberg`, `spark-kubernetes`.
+ Supported configuration classifications:    
[See the AWS documentation website for more details](http://docs.aws.amazon.com/emr/latest/EMR-on-EKS-DevelopmentGuide/emr-eks-6.8.0.html)

  Configuration classifications allow you to customize applications. These often correspond to a configuration XML file for the application, such as `spark-hive-site.xml`. For more information, see [Configure Applications](https://docs.aws.amazon.com/emr/latest/ReleaseGuide/emr-configure-apps.html).

**Notable features**
+ **Spark3.3.0** ‐ Amazon EMR on EKS 6.8 includes Spark 3.3.0, which supports using separate node selector labels for Spark driver executor pods. These new labels enable you to define the node types for the driver and executor pods separately in the StartJobRun API, without using pod templates.
  + Driver node selector property: spark.kubernetes.driver.node.selector.[labelKey]
  + Executor node selector property: spark.kubernetes.executor.node.selector.[labelKey]
+ **Enhanced job failure message** ‐ This release introduces the configuration `spark.stage.extraDetailsOnFetchFailures.enabled` and `spark.stage.extraDetailsOnFetchFailures.maxFailuresToInclude` to track task failures due to user code. These details will be used to enhance the failure message displayed in the driver log when a stage is aborted due to shuffle fetch failure.    
[See the AWS documentation website for more details](http://docs.aws.amazon.com/emr/latest/EMR-on-EKS-DevelopmentGuide/emr-eks-6.8.0.html)

For more information see the [Apache Spark configuration documentation](https://spark.apache.org/docs/latest/running-on-kubernetes.html#configuration).

**Known issue**
+ Amazon EMR on EKS 6.8.0 incorrectly populates the build hash in Parquet files metadata generated using [Apache Spark](https://aws.amazon.com/emr/features/spark/). This issue may cause tools that parse the metadata version string from Parquet files generated by Amazon EMR on EKS 6.8.0 to fail. Customers who parse the version string from Parquet metadata and depend on build hash should switch to a different Amazon EMR version and rewrite the file.

**Resolved issue**
+ **Interrupt Kernel capability for pySpark kernels** ‐ In progress interactive workloads that are triggered by executing cells in a notebook can be stopped by using the `Interrupt Kernel` capability. A fix has been introduced so that this functionality works for pySpark kernels. This is also available in open source at [Changes for handling interrupts for PySpark Kubernetes Kernel \#1115](https://github.com/jupyter-server/enterprise_gateway/pull/1115).