Troubleshooting

Anomaly detection requires at least three published data points for an EvaluationContext before it can score a run. Keep enableDataQualityResultsPublishing set to true from the first run onward.

Rule results don't appear anywhere in Amazon Glue Data Quality. Confirm that enableDataQualityResultsPublishing is set to true in the transform options, and that the role that runs your Amazon EMR Spark application has glue:AccessDataQualityRuntimeConfiguration. Without the permission, the library silently disables publishing while evaluation continues in-process; the transform still returns a result DataFrame but nothing is persisted to Amazon Glue Data Quality.

Results don't appear under my Amazon Glue Data Catalog table in the console. To attribute an Amazon EMR run to a specific catalog table, pass the publishToDataCatalogTarget transform option with DatabaseName and TableName (and optionally CatalogId). This option only takes effect when publishing is already set up as described in the previous bullet. Without this option, runs are grouped only by EvaluationContext.

Runs from Amazon EMR and Amazon Glue are not linked in the console. Both engines must use the same EvaluationContext value. See Comparing runs across Amazon EMR and Amazon Glue.