Troubleshooting errors in Amazon Glue for Spark - Amazon Glue
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Troubleshooting errors in Amazon Glue for Spark

If you encounter errors in Amazon Glue, use the following solutions to help you find the source of the problems and fix them.


The Amazon Glue GitHub repository contains additional troubleshooting guidance in Amazon Glue Frequently Asked Questions.

Error: Resource unavailable

If Amazon Glue returns a resource unavailable message, you can view error messages or logs to help you learn more about the issue. The following tasks describe general methods for troubleshooting.

  • For any connections and development endpoints that you use, check that your cluster has not run out of elastic network interfaces.

Error: Could not find S3 endpoint or NAT gateway for subnetId in VPC

Check the subnet ID and VPC ID in the message to help you diagnose the issue.

  • Check that you have an Amazon S3 VPC endpoint set up, which is required with Amazon Glue. In addition, check your NAT gateway if that's part of your configuration. For more information, see Amazon VPC endpoints for Amazon S3.

Error: Inbound rule in security group required

At least one security group must open all ingress ports. To limit traffic, the source security group in your inbound rule can be restricted to the same security group.

Error: Outbound rule in security group required

At least one security group must open all egress ports. To limit traffic, the source security group in your outbound rule can be restricted to the same security group.

Error: Job run failed because the role passed should be given assume role permissions for the Amazon Glue service

The user who defines a job must have permission for iam:PassRole for Amazon Glue.

Error: DescribeVpcEndpoints action is unauthorized. unable to validate VPC ID vpc-id

  • Check the policy passed to Amazon Glue for the ec2:DescribeVpcEndpoints permission.

Error: DescribeRouteTables action is unauthorized. unable to validate subnet id: Subnet-id in VPC id: vpc-id

  • Check the policy passed to Amazon Glue for the ec2:DescribeRouteTables permission.

Error: Failed to call ec2:DescribeSubnets

  • Check the policy passed to Amazon Glue for the ec2:DescribeSubnets permission.

Error: Failed to call ec2:DescribeSecurityGroups

  • Check the policy passed to Amazon Glue for the ec2:DescribeSecurityGroups permission.

Error: Could not find subnet for AZ

  • The Availability Zone might not be available to Amazon Glue. Create and use a new subnet in a different Availability Zone from the one specified in the message.

Error: Job run exception when writing to a JDBC target

When you are running a job that writes to a JDBC target, the job might encounter errors in the following scenarios:

  • If your job writes to a Microsoft SQL Server table, and the table has columns defined as type Boolean, then the table must be predefined in the SQL Server database. When you define the job on the Amazon Glue console using a SQL Server target with the option Create tables in your data target, don't map any source columns to a target column with data type Boolean. You might encounter an error when the job runs.

    You can avoid the error by doing the following:

    • Choose an existing table with the Boolean column.

    • Edit the ApplyMapping transform and map the Boolean column in the source to a number or string in the target.

    • Edit the ApplyMapping transform to remove the Boolean column from the source.

  • If your job writes to an Oracle table, you might need to adjust the length of names of Oracle objects. In some versions of Oracle, the maximum identifier length is limited to 30 bytes or 128 bytes. This limit affects the table names and column names of Oracle target data stores.

    You can avoid the error by doing the following:

    • Name Oracle target tables within the limit for your version.

    • The default column names are generated from the field names in the data. To handle the case when the column names are longer than the limit, use ApplyMapping or RenameField transforms to change the name of the column to be within the limit.

Error: Amazon S3: The operation is not valid for the object's storage class

If Amazon Glue returns this error, your Amazon Glue job may have been reading data from tables that have partitions across Amazon S3 storage class tiers.

  • By using storage class exclusions, you can ensure that your Amazon Glue jobs will work on tables that have partitions across these storage class tiers. Without exclusions, jobs that read data from these tiers fail with the following error: AmazonS3Exception: The operation is not valid for the object's storage class.

    For more information, see Excluding Amazon S3 storage classes.

Error: Amazon S3 timeout

If Amazon Glue returns a connect timed out error, it might be because it is trying to access an Amazon S3 bucket in another Amazon Region.

  • An Amazon S3 VPC endpoint can only route traffic to buckets within an Amazon Region. If you need to connect to buckets in other Regions, a possible workaround is to use a NAT gateway. For more information, see NAT Gateways.

Error: Amazon S3 access denied

If Amazon Glue returns an access denied error to an Amazon S3 bucket or object, it might be because the IAM role provided does not have a policy with permission to your data store.

  • An ETL job must have access to an Amazon S3 data store used as a source or target. A crawler must have access to an Amazon S3 data store that it crawls. For more information, see Step 2: Create an IAM role for Amazon Glue.

Error: Amazon S3 access key ID does not exist

If Amazon Glue returns an access key ID does not exist error when running a job, it might be because of one of the following reasons:

  • An ETL job uses an IAM role to access data stores, confirm that the IAM role for your job was not deleted before the job started.

  • An IAM role contains permissions to access your data stores, confirm that any attached Amazon S3 policy containing s3:ListBucket is correct.

Error: Job run fails when accessing Amazon S3 with an s3a:// URI

If a job run returns an error like Failed to parse XML document with handler class , it might be because of a failure trying to list hundreds of files using an s3a:// URI. Access your data store using an s3:// URI instead. The following exception trace highlights the errors to look for:

1. com.amazonaws.SdkClientException: Failed to parse XML document with handler class$ListBucketHandler 2. at 3. at 4. at$ListObjectsUnmarshaller.unmarshall( 5. at$ListObjectsUnmarshaller.unmarshall( 6. at 7. at 8. at com.amazonaws.http.response.AwsResponseHandlerAdapter.handle( 9. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.handleResponse( 10. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeOneRequest( 11. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeHelper( 12. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.doExecute( 13. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeWithTimer( 14. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.execute( 15. at com.amazonaws.http.AmazonHttpClient$RequestExecutor.access$500( 16. at com.amazonaws.http.AmazonHttpClient$RequestExecutionBuilderImpl.execute( 17. at com.amazonaws.http.AmazonHttpClient.execute( 18. at 19. at 20. at 21. at 22. at org.apache.hadoop.fs.s3a.S3AFileSystem.getFileStatus( 23. at org.apache.hadoop.fs.s3a.S3AFileSystem.deleteUnnecessaryFakeDirectories( 24. at org.apache.hadoop.fs.s3a.S3AFileSystem.finishedWrite( 25. at org.apache.hadoop.fs.s3a.S3AOutputStream.close( 26. at org.apache.hadoop.fs.FSDataOutputStream$PositionCache.close( 27. at org.apache.hadoop.fs.FSDataOutputStream.close( 28. at org.apache.parquet.hadoop.ParquetFileWriter.end( 29. at org.apache.parquet.hadoop.InternalParquetRecordWriter.close( 30. at org.apache.parquet.hadoop.ParquetRecordWriter.close( 31. at org.apache.spark.sql.execution.datasources.parquet.ParquetOutputWriter.close(ParquetOutputWriter.scala:44) 32. at org.apache.spark.sql.execution.datasources.FileFormatWriter$SingleDirectoryWriteTask.releaseResources(FileFormatWriter.scala:252) 33. at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:191) 34. at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask$3.apply(FileFormatWriter.scala:188) 35. at org.apache.spark.util.Utils$.tryWithSafeFinallyAndFailureCallbacks(Utils.scala:1341) 36. at org.apache.spark.sql.execution.datasources.FileFormatWriter$.org$apache$spark$sql$execution$datasources$FileFormatWriter$$executeTask(FileFormatWriter.scala:193) 37. at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:129) 38. at org.apache.spark.sql.execution.datasources.FileFormatWriter$$anonfun$write$1$$anonfun$3.apply(FileFormatWriter.scala:128) 39. at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:87) 40. at 41. at org.apache.spark.executor.Executor$ 42. at java.util.concurrent.ThreadPoolExecutor.runWorker( 43. at java.util.concurrent.ThreadPoolExecutor$ 44. at

Error: Amazon S3 service token expired

When moving data to and from Amazon Redshift, temporary Amazon S3 credentials, which expire after 1 hour, are used. If you have a long running job, it might fail. For information about how to set up your long running jobs to move data to and from Amazon Redshift, see aws-glue-programming-etl-connect-redshift-home.

Error: No private DNS for network interface found

If a job fails or a development endpoint fails to provision, it might be because of a problem in the network setup.

  • If you are using the Amazon provided DNS, the value of enableDnsHostnames must be set to true. For more information, see DNS .

Error: Development endpoint provisioning failed

If Amazon Glue fails to successfully provision a development endpoint, it might be because of a problem in the network setup.

  • When you define a development endpoint, the VPC, subnet, and security groups are validated to confirm that they meet certain requirements.

  • If you provided the optional SSH public key, check that it is a valid SSH public key.

  • Check in the VPC console that your VPC uses a valid DHCP option set. For more information, see DHCP option sets.

  • If the cluster remains in the PROVISIONING state, contact Amazon Web Services Support.

Error: Notebook server CREATE_FAILED

If Amazon Glue fails to create the notebook server for a development endpoint, it might be because of one of the following problems:

  • Amazon Glue passes an IAM role to Amazon EC2 when it is setting up the notebook server. The IAM role must have a trust relationship to Amazon EC2.

  • The IAM role must have an instance profile of the same name. When you create the role for Amazon EC2 with the IAM console, the instance profile with the same name is automatically created. Check for an error in the log regarding the instance profile name that is not valid. For more information, see Using Instance Profiles .

  • Check that your role has permission to access aws-glue* buckets in the policy that you pass to create the notebook server.

Error: Local notebook fails to start

If your local notebook fails to start and reports errors that a directory or folder cannot be found, it might be because of one of the following problems:

  • If you are running on Microsoft Windows, make sure that the JAVA_HOME environment variable points to the correct Java directory. It's possible to update Java without updating this variable, and if it points to a folder that no longer exists, Jupyter notebooks fail to start.

Error: Running crawler failed

If Amazon Glue fails to successfully run a crawler to catalog your data, it might be because of one of the following reasons. First check if an error is listed in the Amazon Glue console crawlers list. Check if there is an exclamation icon next to the crawler name and hover over the icon to see any associated messages.

  • Check the logs for the crawler run in CloudWatch Logs under /aws-glue/crawlers.

Error: Partitions were not updated

In case your partitions were not updated in the Data Catalog when you ran an ETL job, these log statements from the DataSink class in the CloudWatch logs may be helpful:

  • "Attempting to fast-forward updates to the Catalog - nameSpace:"  —  Shows which database, table, and catalogId are attempted to be modified by this job. If this statement is not here, check if enableUpdateCatalog is set to true and properly passed as a getSink() parameter or in additional_options.

  • "Schema change policy behavior:"  —  Shows which schema updateBehavior value you passed in.

  • "Schemas qualify (schema compare):"  —  Will be true or false.

  • "Schemas qualify (case-insensitive compare):"  —  Will be true or false.

  • If both are false and your updateBehavior is not set to UPDATE_IN_DATABASE, then your DynamicFrame schema needs to be identical or contain a subset of the columns seen in the Data Catalog table schema.

For more information on updating partitions, see Updating the schema, and adding new partitions in the Data Catalog using Amazon Glue ETL jobs.

Error: Job bookmark update failed due to version mismatch

You may be trying to parametize Amazon Glue jobs to apply the same transformation/logic on different datasets in Amazon S3. You want to track processed files on the locations provided. When you run the same job on the same source bucket and write to the same/different destination concurrently (concurrency >1) the job fails with this error:

py4j.protocol.Py4JJavaError: An error occurred while Continuation update failed due to version mismatch. Expected version 2 but found version 3

Solution: set concurrency to 1 or don't run the job concurrently.

Currently Amazon Glue bookmarks don't support concurrent job runs and commits will fail.

Error: A job is reprocessing data when job bookmarks are enabled

There might be cases when you have enabled Amazon Glue job bookmarks, but your ETL job is reprocessing data that was already processed in an earlier run. Check for these common causes of this error:

Max Concurrency

Setting the maximum number of concurrent runs for the job greater than the default value of 1 can interfere with job bookmarks. This can occur when job bookmarks check the last modified time of objects to verify which objects need to be reprocessed. For more information, see the discussion of max concurrency in Configuring job properties for Spark jobs in Amazon Glue.

Missing Job Object

Ensure that your job run script ends with the following commit:


When you include this object, Amazon Glue records the timestamp and path of the job run. If you run the job again with the same path, Amazon Glue processes only the new files. If you don't include this object and job bookmarks are enabled, the job reprocesses the already processed files along with the new files and creates redundancy in the job's target data store.

Missing Transformation Context Parameter

Transformation context is an optional parameter in the GlueContext class, but job bookmarks don't work if you don't include it. To resolve this error, add the transformation context parameter when you create the DynamicFrame, as shown following:

sample_dynF=create_dynamic_frame_from_catalog(database, table_name,transformation_ctx="sample_dynF")
Input Source

If you are using a relational database (a JDBC connection) for the input source, job bookmarks work only if the table's primary keys are in sequential order. Job bookmarks work for new rows, but not for updated rows. That is because job bookmarks look for the primary keys, which already exist. This does not apply if your input source is Amazon Simple Storage Service (Amazon S3).

Last Modified Time

For Amazon S3 input sources, job bookmarks check the last modified time of the objects, rather than the file names, to verify which objects need to be reprocessed. If your input source data has been modified since your last job run, the files are reprocessed when you run the job again.

Error: Failover behavior between VPCs in Amazon Glue

The following process is used for failover for jobs in Amazon Glue 4.0 and previous versions.

Summary: an Amazon Glue connection is selected at the time a job run is submitted. If the job run encounters some issues, (lack of IP addresses, connectivity to source, routing problem), the job run will fail. If retries are configured, Amazon Glue will retry with the same connection.

  1. For each run attempt, Amazon Glue will check the connections health in the order listed in the job configuration, given until it finds one it can use. In the case of an Availability Zone (AZ) failure, the connections from that AZ will fail the check and will be skipped.

  2. Amazon Glue validates the connection with the following:

    • checks for valid Amazon VPC id and subnet.

    • checks that a NAT gateway or Amazon VPC endpoint exists.

    • checks that the subnet has more than 0 allocated IP addresses.

    • checks that the AZ is healthy.

    Amazon Glue cannot verify connectivity at the time of job run submission.

  3. For jobs using Amazon VPC, all drivers and executors will be created in the same AZ with the connection selected at the time of job run submission.

  4. If retries are configured, Amazon Glue will retry with the same connection. This is because we cannot guarantee problems with this connection are long-running. If an AZ fails, existing job runs (depending on the stage of the job run) in that AZ can fail. A retry should detect an AZ failure and choose another AZ for the new run.