apache/druid · error · IllegalStateException

There can only be one aggregation for column

Error message

There can only be one aggregation for column [%s].

What it means

Raised in SegmentGenerationUtils.makeDimensionsAndAggregatorsForIngestion when, for a rollup MSQ ingestion, an output column ends up with more than one AggregatorFactory. Each rollup output column must be produced by exactly one aggregator; duplicates indicate a corrupted columnMappings/aggregator mapping built from the query plan.

Solutions

1. Check that the ingestion SQL does not alias the same aggregated expression to a column name that is aggregated again.
2. If this arises from an internally generated plan, file a bug with the query: the columnMappings → aggregator assignment should never produce duplicates.
3. Work around by renaming conflicting output columns in the SQL.

When it happens

Trigger: Thrown at multi-stage-query/src/main/java/org/apache/druid/msq/indexing/destination/SegmentGenerationUtils.java:363 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/b3a9490e1491b545. Report an issue: GitHub.

Appendix: source

Thrown at multi-stage-query/src/main/java/org/apache/druid/msq/indexing/destination/SegmentGenerationUtils.java:363

    for (final KeyColumn clusterByColumn : queryClusterBy.getColumns()) {
      final IntList outputColumns = columnMappings.getOutputColumnsForQueryColumn(clusterByColumn.columnName());
      for (final int outputColumn : outputColumns) {
        outputColumnsInOrder.add(columnMappings.getOutputColumnName(outputColumn));
      }
    }

    // Then all other columns.
    outputColumnsInOrder.addAll(columnMappings.getOutputColumnNames());

    Map<String, AggregatorFactory> outputColumnAggregatorFactories = new HashMap<>();

    if (isRollupQuery) {
      // Populate aggregators from the native query when doing an ingest in rollup mode.
      for (AggregatorFactory aggregatorFactory : ((GroupByQuery) query).getAggregatorSpecs()) {
        for (final int outputColumn : columnMappings.getOutputColumnsForQueryColumn(aggregatorFactory.getName())) {
          final String outputColumnName = columnMappings.getOutputColumnName(outputColumn);
          if (outputColumnAggregatorFactories.containsKey(outputColumnName)) {
            throw new ISE("There can only be one aggregation for column [%s].", outputColumn);
          } else {
            outputColumnAggregatorFactories.put(
                outputColumnName,
                aggregatorFactory.withName(outputColumnName).getCombiningFactory()
            );
          }
        }
      }
    }

    // Each column can be either a dimension or an aggregator.
    // For non-complex columns, If the aggregator factory of the column is not available, we treat the column as
    // a dimension. For complex columns, certains hacks are in place.
    for (final String outputColumnName : outputColumnsInOrder) {
      // CollectionUtils.getOnlyElement because this method is only called during ingestion, where we require
      // that output names be unique.
      final int outputColumn = CollectionUtils.getOnlyElement(
          columnMappings.getOutputColumnsByName(outputColumnName),

View on GitHub (pinned to 9b90983fd2)