zookeeper-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From hanm <...@git.apache.org>
Subject [GitHub] zookeeper pull request #247: Potential server deadlock between follower sync...
Date Sat, 06 May 2017 05:27:57 GMT
Github user hanm commented on a diff in the pull request:

    --- Diff: src/java/main/org/apache/zookeeper/server/quorum/QuorumPeer.java ---
    @@ -682,27 +682,19 @@ public void setQuorumAddress(InetSocketAddress addr){
         public InetSocketAddress getElectionAddress(){
    -        synchronized (QV_LOCK) {
    -            return myElectionAddr;
    -        }
    +        return myElectionAddr;
    --- End diff --
    Here is why. 
    * All set code path was protected by QV_LOCK already, which implies that whoever calls
set* should already acquire the QV_LOCK.
    * On get code path, it is not guaranteed that caller will hold QV_LOCK while a reconfig
op is in flight, and I think it is possible in theory we could get out dated election address,
e.g. at [this place](https://github.com/apache/zookeeper/blob/master/src/java/main/org/apache/zookeeper/server/quorum/QuorumCnxManager.java#L265).
Hopefully this is not a problem because we don't require precise information being sent here
- if we get out dated addr (in case the current quorum peer is being reconfigured) and sent
this to another peer, another peer will not able to connect but that's fine, it will retry
until at certain point later it will get correct information. 

If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.

View raw message