[
https://issues.apache.org/jira/browse/ZOOKEEPER-1054?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13932531#comment-13932531
]
Alexander Shraer commented on ZOOKEEPER-1054:
---------------------------------------------
the way reconfig works currently is that the leader first accepts connections
and then we invoke a reconfig
so without looking at the patch, if the leader or others (during fle) reject
new servers then join won't work.
we did discuss in the past adding an ensemble id / db_id to the config, and
check that everywhere, so perhaps that could be an alternative
> Drop connections from servers not in the cluster configuration
> --------------------------------------------------------------
>
> Key: ZOOKEEPER-1054
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1054
> Project: ZooKeeper
> Issue Type: Improvement
> Components: leaderElection
> Reporter: Bhallamudi Venkata Siva Kamesh
> Assignee: Bhallamudi Venkata Siva Kamesh
> Priority: Minor
> Labels: security
> Fix For: 3.5.0
>
> Attachments: ZOOKEEPER-1054-1.patch, ZOOKEEPER-1054-2.patch,
> ZOOKEEPER-1054.patch, zookeeper-1054-3.patch, zookeeper-1054-4.patch
>
>
> Let us suppose zookeeper cluster is running in the following machines
> {noformat}
> server.1=10.18.52.133:2999:3999
> server.2=10.18.52.253:2999:3999
> server.3=10.18.52.96:2999:3999
> {noformat}
> Let us take another zookeeper(10.18.52.109),which is not part of the cluster
> configuration, tries to participate in the leader election,then one of the
> zookeeper server's log is filled with following INFO messages
> {noformat}
> 2011-04-19 17:42:42,457 - INFO
> [/10.18.52.133:3999:QuorumCnxManager$Listener@486] - Received connection
> request /10.18.52.109:18324
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.2#6252)
