apache/hadoop · error · IOException

Rename unsuccessful : '%s' to '%s'

Error message

Rename unsuccessful : '%s' to '%s'

What it means

flush() commits by renaming files: the new store is written to <path>_NEW, the current file is renamed to <path>_OLD, and renameOrFail promotes _NEW onto the live path. fs.rename returned false and the provider converts that into an IOException naming both paths. This is a filesystem-level refusal - typically an existing destination, missing permissions, or non-atomic rename semantics - not an HDFS exception.

Source

Thrown at hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/crypto/key/JavaKeyStoreProvider.java:643

      renameOrFail(path, oldPath);
      return true;
    } catch (FileNotFoundException e) {
      return false;
    }
  }

  private void revertFromOld(Path oldPath, boolean fileExisted)
      throws IOException {
    if (fileExisted) {
      renameOrFail(oldPath, path);
    }
  }


  private void renameOrFail(Path src, Path dest)
      throws IOException {
    if (!fs.rename(src, dest)) {
      throw new IOException("Rename unsuccessful : "
          + String.format("'%s' to '%s'", src, dest));
    }
  }

  @Override
  public String toString() {
    return uri.toString();
  }

  /**
   * The factory to create JksProviders, which is used by the ServiceLoader.
   */
  public static class Factory extends KeyProviderFactory {
    @Override
    public KeyProvider createProvider(URI providerName,
                                      Configuration conf) throws IOException {
      if (SCHEME_NAME.equals(providerName.getScheme())) {
        return new JavaKeyStoreProvider(providerName, conf);

View on GitHub (pinned to 2add963021)

Solutions

1. Delete stale <keystore>_NEW and <keystore>_OLD files left by an interrupted flush, then retry
2. Verify write permission on the keystore's parent directory for the running user
3. Ensure exactly one writer per keystore file; serialize flush across processes
4. Keep keystores on filesystems with proper rename support (HDFS or local POSIX FS)

// before flush: clear stale side files from an interrupted run and check dir writability
org.apache.hadoop.fs.Path p = new org.apache.hadoop.fs.Path(providerUri);
FileSystem fs = p.getFileSystem(conf);
fs.delete(new Path(p + "_NEW"), false);
fs.delete(new Path(p + "_OLD"), false);

try { provider.flush(); } catch (IOException e) { if (String.valueOf(e.getMessage()).startsWith("Rename unsuccessful")) { // remove stale _NEW/_OLD, verify parent dir permissions, retry flush once } else { throw e; } }

When it happens

Trigger: A leftover *_NEW or destination file from a crashed flush blocks the rename; no write permission on the keystore's parent directory; two processes flushing the same keystore concurrently; a filesystem (NFS, object-store mount) without POSIX rename guarantees.

Common situations: Two daemons sharing one keystore file; keystore directory made read-only; keystore stored on a mount with unusual rename semantics; interrupted process leaving side files behind.

Related errors

• Can't set metadata key ${entry.getKey()}
• Can't store keystore ${this}
• Cannot overwrite an existing file: %s
• Cannot rename because path does not exist: %s
• Wrong key length. Required ${options.getBitLength()}, but go

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/12bd37474755f1a8. Report an issue: GitHub.