apache/druid · error · ResponseException

error closing org.apache.druid.query.aggregation.Serializabl

Error message

error closing org.apache.druid.query.aggregation.SerializablePairLongFloatComplexColumn

What it means

SerializablePairLongFloatComplexColumn.close() delegates cleanup to a Closer and wraps any IOException thrown during close in a Druid RE with 'error closing <class name>'. It guards against silent resource leaks when a serialized pair-long-float column cannot be cleanly released.

Source

Thrown at processing/src/main/java/org/apache/druid/query/aggregation/SerializablePairLongFloatComplexColumn.java:86

  public Object getRowValue(int rowNum)
  {
    return serde.deserialize(cellReader.getCell(rowNum));
  }

  @Override
  public int getLength()
  {
    return serializedSize;
  }

  @Override
  public void close()
  {
    try {
      closer.close();
    }
    catch (IOException e) {
      throw new RE(e, "error closing " + getClass().getName());
    }
  }

  public static class Builder
  {
    private final int serializedSize;
    private final AbstractSerializablePairLongObjectDeltaEncodedStagedSerde<?> serde;
    private final CellReader.Builder cellReaderBuilder;

    public Builder(ByteBuffer buffer)
    {
      ByteBuffer masterByteBuffer = buffer.asReadOnlyBuffer().order(ByteOrder.nativeOrder());

      serializedSize = masterByteBuffer.remaining();

      AbstractSerializablePairLongObjectColumnHeader<?> columnHeader =
          AbstractSerializablePairLongObjectColumnHeader.fromBuffer(masterByteBuffer, SerializablePairLongFloat.class);

View on GitHub (pinned to 9b90983fd2)

Solutions

Look at the IOException cause to identify which underlying resource failed to close and fix that I/O condition
Avoid closing the underlying channel independently of the Closer (prevents double-close failures)
Ensure segment lifecycle ownership is clear — only one component closes the column
Check filesystem health/permissions on the segment storage location

Example fix

// before
closer.close(); // IOException propagates as RE
// after
try {
  closer.close();
} catch (IOException e) {
  LOG.debug(e, "ignoring close failure for %s", getClass().getName());
}

Defensive patterns

Strategy: try-catch

Try / catch

try {
  column.close();
} catch (RE e) {
  if (e.getCause() instanceof IOException) {
    LOG.warn(e, "failed to close SerializablePairLongFloatComplexColumn");
  } else {
    throw e;
  }
}

Prevention

Register all underlying Closeables once with the Closer and let it manage ordering
Avoid races between segment cleanup and column close (single lifecycle owner)
Check filesystem health and permissions on segment directories
Watch logs for repeated close failures pointing at leaked descriptors

When it happens

Trigger: close() is invoked and the underlying Closeable (e.g., file channel or mapped byte buffer resource) throws IOException during close.

Common situations: Segment files removed or already closed by another component before column cleanup; filesystem/permission errors; double-close of channels registered in the Closer; broker/historical task shutdown ordering causing races on segment files.

Related errors

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/8544f32b5d869b9a. Report an issue: GitHub.