apache/hadoop · error · java.io.IOException

Cannot rollback. None of the storage directories contain pre

Error message

Cannot rollback. None of the storage directories contain previous fs state.

What it means

doRollback() requires a previous/ snapshot in at least one local storage directory — or, in HA, a shared edits dir that can roll back. Finding none means there is nothing to revert to: typically the upgrade was already finalized (finalize removes previous/) or this namespace never upgraded, so -rollback aborts.

Source

Thrown at hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSImage.java:550

          continue;
        }
        LOG.info("Can perform rollback for " + sd);
        canRollback = true;
      }
      
      if (fsns.isHaEnabled()) {
        // If HA is enabled, check if the shared log can be rolled back as well.
        editLog.initJournalsForWrite();
        boolean canRollBackSharedEditLog = editLog.canRollBackSharedLog(
            prevState.getStorage(), storage.getServiceLayoutVersion());
        if (canRollBackSharedEditLog) {
          LOG.info("Can perform rollback for shared edit log.");
          canRollback = true;
        }
      }
      
      if (!canRollback)
        throw new IOException("Cannot rollback. None of the storage "
            + "directories contain previous fs state.");
  
      // Now that we know all directories are going to be consistent
      // Do rollback for each directory containing previous state
      for (Iterator<StorageDirectory> it = storage.dirIterator(false); it.hasNext();) {
        StorageDirectory sd = it.next();
        LOG.info("Rolling back storage directory " + sd.getRoot()
                 + ".\n   new LV = " + prevState.getStorage().getLayoutVersion()
                 + "; new CTime = " + prevState.getStorage().getCTime());
        NNUpgradeUtil.doRollBack(sd);
      }
      if (fsns.isHaEnabled()) {
        // If HA is enabled, try to roll back the shared log as well.
        editLog.doRollback();
      }
      
      isUpgradeFinalized = true;
    } finally {

View on GitHub (pinned to 2add963021)

Solutions

1. Verify whether previous/ exists: ls <each dfs.namenode.name.dir>/previous.
2. If it exists on an unmounted disk, fix mounts and retry.
3. If the upgrade was already finalized, rollback is impossible: stay on the current version, downgrade binaries if the layout still matches, or restore metadata from backup.
4. In HA, also verify the shared edits dir (JournalNodes) can roll back.

Example fix

# before
hdfs namenode -rollback
# IOException: Cannot rollback. None of the storage directories contain previous fs state.

# after: confirm state, then choose another path
ls /dfs/nn/current /dfs/nn/previous 2>&1   # previous/ absent => already finalized
hdfs dfsadmin -fs nn:8020 -report           # proceed on current version, or restore backup

boolean anyPrevious = nameDirs.stream()
    .anyMatch(u -> Files.exists(Paths.get(u.getPath(), "previous")));
if (!anyPrevious) {
  throw new IllegalStateException(
      "Nothing to roll back to: no previous/ in any name dir");
}

When it happens

Trigger: Running 'hdfs namenode -rollback' after 'hdfs dfsadmin -finalizeUpgrade' removed previous/; rolling back a freshly formatted namespace; previous/ present on a volume that is not mounted so it is invisible.

Common situations: Operator decides to roll back after finalize already ran; confusion between rollback (needs previous/) and downgrade (needs old binaries); mount lost on the disk holding previous/.

Related errors

• Directory {dir} is in an inconsistent state: previous fs sta
• Cannot rollback to storage version {prevStorage.getLayoutVer
• too many temporary directories.
• *********** Upgrade is not supported from this older versio
• Cannot rollback to a newer state. Datanode previous state: L

AI-assisted analysis of apache/hadoop@2add963021 (2026-08-22). Data as JSON: /api/errors/47bf6e96067d6d85. Report an issue: GitHub.