apache/hadoop · error · UnsupportedOperationException

in.getClass().getName() + ": claims unbuffer capabilty but d

Error message

in.getClass().getName() + ": claims unbuffer capabilty but does not implement CanUnbuffer"

What it means

Thrown by StreamCapabilitiesPolicy.unbuffer when a stream answers true to hasCapability("unbuffer") but the subsequent cast to CanUnbuffer fails with ClassCastException, which the policy converts to UnsupportedOperationException with the class name prefixed. It is a contract violation inside the stream implementation: the stream advertises an unbuffer capability it does not actually implement. The message constant contains the upstream typo 'capabilty'.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/fs/StreamCapabilitiesPolicy.java:55

          StreamCapabilitiesPolicy.class);
  /**
   * Implement the policy for {@link CanUnbuffer#unbuffer()}.
   *
   * @param in the input stream
   */
  public static void unbuffer(InputStream in) {
    try {
      if (in instanceof StreamCapabilities
          && ((StreamCapabilities) in).hasCapability(
          StreamCapabilities.UNBUFFER)) {
        ((CanUnbuffer) in).unbuffer();
      } else {
        LOG.debug(in.getClass().getName() + ":"
                + " does not implement StreamCapabilities"
                + " and the unbuffer capability");
      }
    } catch (ClassCastException e) {
      throw new UnsupportedOperationException(in.getClass().getName() + ": "
              + CAN_UNBUFFER_NOT_IMPLEMENTED_MESSAGE);
    }
  }
}

View on GitHub (pinned to 2add963021)

Solutions

1. Fix the stream class: implement CanUnbuffer and provide a real unbuffer() that releases read-ahead buffers
2. Until implemented, make hasCapability("unbuffer") return false so callers skip the path gracefully
3. As a caller, pre-check 'in instanceof CanUnbuffer' before invoking unbuffer logic

Example fix

// before
public class MyInputStream extends InputStream
    implements StreamCapabilities {
  public boolean hasCapability(String c) {
    return StreamCapabilities.UNBUFFER.equals(c); // promise without implementation
  }
}

// after
public class MyInputStream extends InputStream
    implements StreamCapabilities, CanUnbuffer {
  public boolean hasCapability(String c) {
    return StreamCapabilities.UNBUFFER.equals(c);
  }
  public void unbuffer() { /* release read-ahead buffers */ }
}

public static boolean canUnbuffer(java.io.InputStream in) {
  return in instanceof org.apache.hadoop.fs.CanUnbuffer
      && in instanceof org.apache.hadoop.fs.StreamCapabilities
      && ((org.apache.hadoop.fs.StreamCapabilities) in)
          .hasCapability(org.apache.hadoop.fs.StreamCapabilities.UNBUFFER);
}

// usage: if (canUnbuffer(in)) ((CanUnbuffer) in).unbuffer(); else LOG.debug("no unbuffer");

try {
  StreamCapabilitiesPolicy.unbuffer(in);
} catch (UnsupportedOperationException e) {
  // message: "<Class>: claims unbuffer capabilty but does not implement CanUnbuffer"
  // -> bug in the stream implementation; fix the stream class, do not swallow
  LOG.error("Stream advertises unbuffer but cannot unbuffer: {}", in.getClass(), e);
}

When it happens

Trigger: A custom InputStream implements StreamCapabilities.hasCapability() returning true for StreamCapabilities.UNBUFFER but does not implement the CanUnbuffer interface; unbuffer() is then invoked on it (e.g. via IOUtils/Shell 'unbuffer' paths or FileSystem reopen logic).

Common situations: Third-party or in-house FileSystem connectors that copied a hasCapability table without wiring the corresponding interfaces; upgrading a connector whose capability list gained 'unbuffer' prematurely.

Related errors

• Path must be absolute: " + path
• Append is not supported by BaiduBosFileSystem
• No more entry in " + f
• f + " is a directory"
• f + " already exists"

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/2a5b2d89ea8c1b92. Report an issue: GitHub.