[
https://issues.apache.org/jira/browse/ZOOKEEPER-975?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13004273#comment-13004273
]
Vishal K commented on ZOOKEEPER-975:
------------------------------------
Hi Flavio,
I have a patch for this, but I have it on the top of the fix for ZOOKEEPER-932.
We have 932 applied to our ZK code since we need it. Until ZOOKEEPER-932 is
reviewed and committed, I will have to keep back porting patches (and do double
testing). I will port my changes to trunk if someone requires a fix for the
bug. Since this is not a blocker, I am going to hold off the patch until 932 is
reviewed. That will reduce my testing and porting overhead. Does that sound ok?
The patch I have is good only for FLE.
{quote}
About maintenance, we have some time back talked about maintaining only the TCP
version of FLE (FLE+QCM). There was never some real pressure to eliminate the
others, and in fact previously some users were still using LE. I'm all for
maintaining only FLE, but we need to hear the opinion of others. More thoughts?
{quote}
The documentation says: "The implementations of leader election 1 and 2 are
currently not supported, and we have the intention of deprecating them in the
near future. Implementations 0 and 3 are currently supported, and we plan to
keep supporting them in the near future. To avoid having to support multiple
versions of leader election unecessarily, we may eventually consider
deprecating algorithm 0 as well, but we will plan according to the needs of the
community."
Is there a significant advantage of using LE 0 vs LE 3?
> new peer goes in LEADING state even if ensemble is online
> ---------------------------------------------------------
>
> Key: ZOOKEEPER-975
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-975
> Project: ZooKeeper
> Issue Type: Bug
> Affects Versions: 3.3.2
> Reporter: Vishal K
> Fix For: 3.4.0
>
> Attachments: ZOOKEEPER-975.patch
>
>
> Scenario:
> 1. 2 of the 3 ZK nodes are online
> 2. Third node is attempting to join
> 3. Third node unnecessarily goes in "LEADING" state
> 4. Then third goes back to LOOKING (no majority of followers) and finally
> goes to FOLLOWING state.
> While going through the logs I noticed that a peer C that is trying to
> join an already formed cluster goes in LEADING state. This is because
> QuorumCnxManager of A and B sends the entire history of notification
> messages to C. C receives the notification messages that were
> exchanged between A and B when they were forming the cluster.
> In FastLeaderElection.lookForLeader(), due to the following piece of
> code, C quits lookForLeader assuming that it is supposed to lead.
> 740 //If have received from all nodes, then
> terminate
> 741 if ((self.getVotingView().size() ==
> recvset.size()) &&
> 742
> (self.getQuorumVerifier().getWeight(proposedLeader) != 0)){
> 743 self.setPeerState((proposedLeader ==
> self.getId()) ?
> 744 ServerState.LEADING:
> learningState());
> 745 leaveInstance();
> 746 return new Vote(proposedLeader,
> proposedZxid);
> 747
> 748 } else if (termPredicate(recvset,
> This can cause:
> 1. C to unnecessarily go in LEADING state and wait for tickTime * initLimit
> and then restart the FLE.
> 2. C waits for 200 ms (finalizeWait) and then considers whatever
> notifications it has received to make a decision. C could potentially
> decide to follow an old leader, fail to connect to the leader, and
> then restart FLE. See code below.
> 752 if (termPredicate(recvset,
> 753 new Vote(proposedLeader, proposedZxid,
> 754 logicalclock))) {
> 755
> 756 // Verify if there is any change in the
> proposed leader
> 757 while((n = recvqueue.poll(finalizeWait,
> 758 TimeUnit.MILLISECONDS)) != null){
> 759 if(totalOrderPredicate(n.leader,
> n.zxid,
> 760 proposedLeader,
> proposedZxid)){
> 761 recvqueue.put(n);
> 762 break;
> 763 }
> 764 }
> In general, this does not affect correctness of FLE since C will
> eventually go back to FOLLOWING state (A and B won't vote for
> C). However, this delays C from joining the cluster. This can in turn
> affect recovery time of an application.
> Proposal: A and B should send only the latest notification (most
> recent) instead of the entire history. Does this sound reasonable?
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
