apache/druid · error · IllegalArgumentException

Invalid table size[ ]

Error message

Invalid table size[%s]

What it means

TableLongEncodingReader decodes a table-encoded long column from a ByteBuffer. The header's tableSize must be between 0 and CompressionFactory.MAX_TABLE_SIZE; anything else indicates corrupt or malformed segment data, so the constructor fails fast with this IAE instead of allocating an unbounded array.

Solutions

1. Re-download / reload the affected segment (delete from segment cache so it is re-fetched)
2. Verify segment integrity and re-ingest the datasource if corruption persists
3. Check the buffer alignment/version of whatever produced the ByteBuffer being read
4. Confirm the writer and reader Druid versions are compatible

Example fix

// before
ColumnHolder holder = segmentLoader.getColumnator().read(file); // throws on corrupt segment
// after
// validate & re-fetch the segment:
segmentManager.dropSegment(descriptor);
// then allow the loader to re-download a fresh copy before retrying reads

// validate the header before constructing:
int tableSize = buf.duplicate().position(pos+1).getInt();
if (tableSize < 0 || tableSize > CompressionFactory.MAX_TABLE_SIZE) { reloadSegment(); }

try { new TableLongEncodingReader(buf); } catch (IAE e) { log.error("corrupt table-encoded column, reloading segment"); reloadAndRetry(); }

When it happens

Trigger: Reading a column whose serialized header contains a tableSize < 0 or > MAX_TABLE_SIZE, typically from a corrupted segment file, truncated/misaligned buffer, or data written by an incompatible writer.

Common situations: Segments corrupted on disk or in deep storage; hand-editing or mis-parsing column buffers; loading a foreign/garbage file as a Druid segment.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

• Metric[ ] loaded up metric[ ] from disk. File names do…
• Actual Row count mismatch. Expected
• Attempt to add row to swapped-out sink for segment
• Cannot read frame of size [%,d] bytes
• Column capacity exceeded

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/00586178c8c7ef45. Report an issue: GitHub.

Appendix: source

Thrown at processing/src/main/java/org/apache/druid/segment/data/TableLongEncodingReader.java:40

import org.apache.druid.java.util.common.IAE;

import java.nio.ByteBuffer;

public class TableLongEncodingReader implements CompressionFactory.LongEncodingReader
{
  private final long[] table;
  private final int bitsPerValue;
  private final ByteBuffer buffer;
  private VSizeLongSerde.LongDeserializer deserializer;

  public TableLongEncodingReader(ByteBuffer fromBuffer)
  {
    this.buffer = fromBuffer.asReadOnlyBuffer();
    byte version = buffer.get();
    if (version == CompressionFactory.TABLE_ENCODING_VERSION) {
      int tableSize = buffer.getInt();
      if (tableSize < 0 || tableSize > CompressionFactory.MAX_TABLE_SIZE) {
        throw new IAE("Invalid table size[%s]", tableSize);
      }
      bitsPerValue = VSizeLongSerde.getBitsForMax(tableSize);
      table = new long[tableSize];
      for (int i = 0; i < tableSize; i++) {
        table[i] = buffer.getLong();
      }
      fromBuffer.position(buffer.position());
      deserializer = VSizeLongSerde.getDeserializer(bitsPerValue, buffer, buffer.position());
    } else {
      throw new IAE("Unknown version[%s]", version);
    }
  }

  private TableLongEncodingReader(ByteBuffer buffer, long[] table, int bitsPerValue)
  {
    this.buffer = buffer;
    this.table = table;
    this.bitsPerValue = bitsPerValue;

View on GitHub (pinned to 9b90983fd2)