apache/hadoop · error · InvalidObjectException

No type in deserialized FileStatus

Error message

No type in deserialized FileStatus

What it means

The second check in FileStatus.validateObject(): after Java-native deserialization, the isdir field (file-vs-directory flag) must be non-null or InvalidObjectException('No type in deserialized FileStatus') is thrown. A typeless FileStatus cannot answer isDirectory(), so the object is rejected rather than allowed to fail later.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/FileStatus.java:545

   * @see PBHelper#convert(FileStatus)
   * @deprecated Use the {@link PBHelper} and protobuf serialization directly.
   */
  @Override
  @Deprecated
  public void write(DataOutput out) throws IOException {
    FileStatusProto proto = PBHelper.convert(this);
    int size = proto.getSerializedSize();
    out.writeInt(size);
    out.write(proto.toByteArray());
  }

  @Override
  public void validateObject() throws InvalidObjectException {
    if (null == path) {
      throw new InvalidObjectException("No Path in deserialized FileStatus");
    }
    if (null == isdir) {
      throw new InvalidObjectException("No type in deserialized FileStatus");
    }
  }



}

View on GitHub (pinned to 2add963021)

Solutions

1. Replace Java serialization of FileStatus with path transfer + re-stat on the receiver
2. Pin identical hadoop-common versions on both JVMs that exchange objects
3. Use the protobuf Writable format (write/readFields) for binary transport instead

Example fix

// before
objectOut.writeObject(fileStatus);
FileStatus back = (FileStatus) objectIn.readObject(); // InvalidObjectException: no type

// after: ship the path, re-stat remotely
objectOut.writeUTF(fileStatus.getPath().toString());
// receiver:
FileStatus back = fileSystem.getFileStatus(new Path(objectIn.readUTF()));

try {
  FileStatus st = (FileStatus) objectIn.readObject();
} catch (InvalidObjectException e) {
  // null isdir after deserialization - class shape mismatch; re-stat instead
  FileStatus st = fileSystem.getFileStatus(knownPath);
}

When it happens

Trigger: ObjectInputStream.readObject() of FileStatus bytes from an incompatible class version where the isdir field did not survive, or a bare no-arg FileStatus serialized before any field was populated.

Common situations: Cross-version object shipping (Spark caching, replicated sessions) where one side runs an older/newer hadoop-common whose FileStatus fields differ.

Related errors

• No Path in deserialized FileStatus
• %s not found: %s
• Can't read FileStatusProto with negative size of ${size}
• Path must be absolute: " + path
• Append is not supported by BaiduBosFileSystem

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/3592fabe9e351f22. Report an issue: GitHub.