[HBASE-9157] ZKUtil.blockUntilAvailable loops forever with non-recoverable errors

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.98.8, 0.99.2
• Component/s: Zookeeper
• Hadoop Flags: Reviewed
• Assignee: Jeffrey Zhong
• Reporter: Jeffrey Zhong
• Created: 08/Aug/13 00:27
• Updated: 21/Feb/15 23:46
• Resolved: 15/Oct/14 18:54

Description

In one of integration test, I observed that a thread keeps spinning error logs "Unexpected exception handling blockUntilAvailable" due to KeeperException.ConnectionLossException. Below is the related code:

    
    while (!finished) {
      try {
        data = ZKUtil.getData(zkw, znode);
      } catch(KeeperException e) {
        LOG.warn("Unexpected exception handling blockUntilAvailable", e);
      }

      if (data == null && (System.currentTimeMillis() +
        HConstants.SOCKET_RETRY_WAIT_MS < endTime)) {
        Thread.sleep(HConstants.SOCKET_RETRY_WAIT_MS);
      } else {
        finished = true;
      }
    }

ConnectionLossException might be recoverable but SessionExpiredException and AuthFailed are not recoverable errors, the while loop can't break.