Mailing-List: contact dev-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@hbase.apache.org
Date: Wed, 15 Jan 2014 09:47:19 +0000 (UTC)
From: "chendihao (JIRA)" <jira@apache.org>
To: dev@hbase.apache.org
Message-ID: <JIRA.12688978.1389779219710.4980.1389779239574@arcas>
In-Reply-To: <JIRA.12688978.1389779219710@arcas>
References: <JIRA.12688978.1389779219710@arcas>
Subject: [jira] [Created] (HBASE-10345) HMaster should not serve when
 disconnected with ZooKeeper
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit

chendihao created HBASE-10345:
---------------------------------

             Summary: HMaster should not serve when disconnected with ZooKeeper
                 Key: HBASE-10345
                 URL: https://issues.apache.org/jira/browse/HBASE-10345
             Project: HBase
          Issue Type: Bug
          Components: master
    Affects Versions: 0.94.3
            Reporter: chendihao


Refer to HBASE-9468(Previous active master can still serves RPC request when it is trying recovering expired zk session), we can fail fast to avoid existing double masters at the same time. But this problem may occur before session expired. When receive Disconnected event, we can't make sure of that this active master can communicate with zk later. And it doesn't know whether backup master has become the new active master or not until it receives Expired event(which may lose forever). During this unsure-who-is-active-master period, the current active master should not serve(maybe turn off RpcServer).

Here is the statement from "ZooKeeper Distributed Process Coordination" P101
{quote}
If the developer is not careful, the old leader will continue to act as a leader and may take actions that conflict with those of the new leader. For this reason, when a process receives a Disconnected event, the process should suspend actions taken as a leader until it reconnects. Normally this reconnect happens very quickly.
{quote}

So it's equally necessary to handle Disconnected event and Expired event.


--
This message was sent by Atlassian JIRA
(v6.1.5#6160)