apache/hadoop · error · IOException

bytes array length error. Actual length is {}

Error message

bytes array length error. Actual length is {}

What it means

In LevelDB mode, the value stored for each directory child in dirChildMap is the 8-byte parent inode id; this decode found a different length. The map is built in-process from the fsimage moments earlier, so a wrong length means the LevelDB store itself got corrupted mid-run (failing disk, another process touching the temp dir) — it is not caused by oiv arguments or the fsimage content.

Source

Thrown at hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/tools/offlineImageViewer/PBImageTextWriter.java:455

      dirChildMap.put(toBytes(childId), toBytes(parentId));
    }

    @Override
    public void putDir(INode dir) throws IOException {
      Preconditions.checkArgument(dir.hasDirectory(),
          "INode %s (%s) is not a directory.", dir.getId(), dir.getName());
      dirMap.put(toBytes(dir.getId()), toBytes(dir.getName().toStringUtf8()));
    }

    private long getFromDirChildMap(long inode) throws IOException {
      byte[] bytes = dirChildMap.get(toBytes(inode));
      if (bytes == null) {
        // The inode is an INodeReference, which is generated from snapshot.
        // For delimited oiv tool, no need to print out metadata in snapshots.
        throw PBImageTextWriter.createIgnoredSnapshotException(inode);
      }
      if (bytes.length != 8) {
        throw new IOException(
            "bytes array length error. Actual length is " + bytes.length);
      }
      return toLong(bytes);
    }

    @Override
    public String getParentPath(long inode) throws IOException {
      if (inode == INodeId.ROOT_INODE_ID) {
        return "/";
      }
      long parent = getFromDirChildMap(inode);
      byte[] bytes = dirMap.get(toBytes(parent));
      synchronized (this) {
        if (!dirPathCache.containsKey(parent)) {
          if (parent != INodeId.ROOT_INODE_ID && bytes == null) {
            // The parent is an INodeReference, which is generated from snapshot.
            // For delimited oiv tool, no need to print out metadata in snapshots.
            throw PBImageTextWriter.createIgnoredSnapshotException(inode);

View on GitHub (pinned to 2add963021)

Solutions

1. Rerun with a brand-new -t directory (delete the old one first)
2. Rerun without -t to keep the metadata map in memory; this isolates the temp storage as the culprit
3. If it reproduces deterministically with a fresh dir and the same fsimage, verify the fsimage md5 and file a HADOOP JIRA for PBImageTextWriter

try {
  // oiv run with -t
} catch (IOException e) {
  if (e.getMessage().contains("bytes array length error")) {
    // internal LevelDB corruption: wipe temp dir and retry once, in memory if possible
    FileUtils.deleteQuietly(new File(tempPath));
    // rerun oiv without -t (or with a fresh -t)
  } else {
    throw e;
  }
}

When it happens

Trigger: Corruption of the LevelDB files under the -t directory while oiv is running: failing disk, another process writing into the live temp dir, or filesystem-level data damage on local or NFS-backed temp storage.

Common situations: Flaky local/NFS temp storage; operators clearing or copying files inside the live temp directory; very large images where long run times increase exposure.

Related errors

• Folder {} already exists! Delete manually or provide another
• Expected to parse {} in parallel, but parsed {}. The image m
• InMemoryAliasMap location is null
• Unable to create aliasmap snapshot directory {newLevelDBDir}
• File " + url + " computed digest " + computedDigest + " does

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/aae9a4fdbe11f625. Report an issue: GitHub.