apache/hadoop · error · IOException

Recorded block size is {}, but datanode returned {} bytes

Error message

Recorded block size is {}, but datanode returned {} bytes

What it means

Error "Recorded block size is {}, but datanode returned {} bytes" thrown in apache/hadoop.

Source

Thrown at hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/NamenodeFsck.java:1149

      }  catch (IOException ex) {
        // Put chosen node into dead list, continue
        LOG.info("Failed to connect to " + targetAddr + ":" + ex);
        deadNodes.add(chosenNode);
      }
    }

    long bytesRead = 0L;
    try {
      bytesRead = copyBock(blockReader, fos);
    } catch (Exception e) {
      throw new Exception("Could not copy block data for " + lblock.getBlock(),
          e);
    } finally {
      blockReader.close();
    }

    if (bytesRead != block.getNumBytes()) {
      throw new IOException("Recorded block size is " + block.getNumBytes()
          + ", but datanode returned " + bytesRead + " bytes");
    }
  }

  private long copyBock(BlockReader blockReader, OutputStream os)
      throws IOException {
    final byte[] buf = new byte[8192];
    int cnt = 0;
    long bytesRead = 0L;
    while ((cnt = blockReader.read(buf, 0, buf.length)) > 0) {
      os.write(buf, 0, cnt);
      bytesRead += cnt;
    }
    return bytesRead;
  }

  @Override
  public DataEncryptionKey newDataEncryptionKey() throws IOException {

View on GitHub (pinned to 2add963021)

Solutions

1. Investigate the DataNode serving the mismatched size; the block may be corrupt — consider recovering or deleting the file.

When it happens

Trigger: A DataNode returns a byte count different from the block size recorded in the NameNode metadata.

Common situations: Disk corruption, truncated blocks, or mixed-version metadata after an unclean shutdown.

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/b6a4ac84d673e1e5. Report an issue: GitHub.