[
https://issues.apache.org/jira/browse/NIFI-2292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15392193#comment-15392193
]
ASF subversion and git services commented on NIFI-2292:
-------------------------------------------------------
Commit 35ff0975bf1193a1c787af365e15a4e0da970ed9 in nifi's branch
refs/heads/master from [~markap14]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=35ff097 ]
NIFI-2292: Funnel all cluster node status changes through the cluster
coordinator instead of having each node broadcast changes to the whole cluster.
This gives us the ability to increment the updateId consistently without race
conditions.
> Nodes in cluster sometimes become out-of-sync with actual 'connection state'
> of node
> ------------------------------------------------------------------------------------
>
> Key: NIFI-2292
> URL: https://issues.apache.org/jira/browse/NIFI-2292
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Mark Payne
> Assignee: Mark Payne
> Fix For: 1.0.0
>
>
> Occasionally I'll see a node that has a different view of the cluster than
> other nodes. Right now I'm actually seeing "node 1" think it's in
> 'CONNECTING' state while nodes 2-5 think we have 5/5 nodes connected.
> This also can result in a node that is elected cluster coordinator and then
> has that role revoked can continually monitor for heartbeats, even though it
> won't receive them since it's not the coordinator anymore. This results in
> continually logging a message like "Failed to retrieve any new heartbeat
> information for nodes. Will not make any decisions based on heartbeats."
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
