apache/druid · error · IllegalArgumentException

Cannot have a comma in the name of a file, got

Error message

Cannot have a comma in the name of a file, got[%s].

What it means

FileSmoosher writes multiple named buffers into a Smoosh 'meta.smoosh' container file, and it uses commas in the internally written file list to delimit entries. Because of that, any file name containing a comma would corrupt the container's metadata, so add() rejects such names up front with this IllegalArgumentException.

Solutions

1. Sanitize the file name before calling add(): replace or strip commas (e.g. name.replace(',', '_')).
2. Ensure upstream segment identifiers/data source names cannot contain commas (validate at ingestion config time).
3. If many parts must be encoded in a name, use a delimiter that is legal (e.g. '_' or '-').

Example fix

// before
smoosher.add(fileName, buffer); // fileName = "ds1,ds2_2020-01-01"
// after
String safeName = fileName.replace(',', '_');
smoosher.add(safeName, buffer);

if (name.indexOf(',') >= 0) {
  throw new IllegalArgumentException("File name must not contain a comma: " + name);
}
smoosher.add(name, buffer);

static boolean isSmooshSafeName(final String name) {
  return name != null && !name.contains(",");
}

try {
  smoosher.add(name, buffer);
} catch (IllegalArgumentException e) {
  throw new IOException("Invalid smoosh file name: " + name, e);
}

When it happens

Trigger: Calling FileSmoosher.add(String name, ByteBuffer bufferToAdd) (directly or via addWithChannel/write) with a name containing a ',' character, e.g. a segment identifier or generated file name that embeds a comma-separated list.

Common situations: Building segment files where the segment id or data source name contains a comma; concatenating dimension values or partition keys into a file name without sanitizing; copying names from CSV input.

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

• Cannot add files of the same name, already have
• A-Not-B requires at least 1 sketch
• BloomKFilters are not compatible for merging. this
• Bucket index out of range (0, )
• Caching is not supported. Check `isCacheable` before…

AI-assisted analysis of apache/druid@9b90983fd2 (2026-09-07). Data as JSON: /api/errors/e4860bd17d8d0654. Report an issue: GitHub.

Appendix: source

Thrown at processing/src/main/java/org/apache/druid/java/util/common/io/smoosh/FileSmoosher.java:146

  @Override
  public void addColumn(String name, ColumnDescriptor columnDescriptor)
  {
    throw DruidException.defensive("not supported");
  }

  @Override
  public void add(String name, File fileToAdd) throws IOException
  {
    try (MappedByteBufferHandler fileMappingHandler = FileUtils.map(fileToAdd)) {
      add(name, fileMappingHandler.get());
    }
  }

  @Override
  public void add(String name, ByteBuffer bufferToAdd) throws IOException
  {
    if (name.contains(",")) {
      throw new IAE("Cannot have a comma in the name of a file, got[%s].", name);
    }

    if (internalFiles.get(name) != null) {
      throw new IAE("Cannot add files of the same name, already have [%s]", name);
    }

    long size = 0;
    size += bufferToAdd.remaining();

    try (SegmentFileChannel out = addWithChannel(name, size)) {
      out.write(bufferToAdd);
    }
  }

  @Override
  public SegmentFileChannel addWithChannel(final String name, final long size) throws IOException
  {
    return addWithSmooshedWriter(name, size);

View on GitHub (pinned to 9b90983fd2)