[
https://issues.apache.org/jira/browse/NIFI-1966?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15468463#comment-15468463
]
ASF subversion and git services commented on NIFI-1966:
-------------------------------------------------------
Commit a7e76cc00a13815d899267ce9e4ca1ad6cf51ce1 in nifi's branch
refs/heads/master from [~markap14]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=a7e76cc ]
NIFI-1966: When cluster is started up, do not assume that Cluster Coordinator
has the golden copy of the flow but instead wait for some period of time or
until the required number of nodes have connected, and then choose which flow
is correct. This closes #977
> Address Issue of Nodes Disagreeing About Flow Contents on Startup
> -----------------------------------------------------------------
>
> Key: NIFI-1966
> URL: https://issues.apache.org/jira/browse/NIFI-1966
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Matt Gilman
> Assignee: Mark Payne
> Priority: Blocker
> Fix For: 1.1.0
>
>
> During startup a coordinator is selected at random to act as the manager of
> data flow and distributing it to other nodes. This works fine unless that
> selected coordinator happens to have a uninheritable data flow. When this
> occurs, the other nodes either are
> - Able to connect to Coordinator and then unable to join due to an
> uninheritable flow exception due to being presented with a bad data flow. In
> this case, the node shut down.
> - Unable to connect to the Coordinator due to lack of quorem and never
> present with the bad data flow. In this case, the node ran its local flow.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
