[
https://issues.apache.org/jira/browse/NIFI-5745?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16664749#comment-16664749
]
ASF subversion and git services commented on NIFI-5745:
-------------------------------------------------------
Commit 234ddb0fe1a36ad947c340114058d82c777d791f in nifi's branch
refs/heads/master from [~markap14]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=234ddb0 ]
NIFI-5745: When determining if backpressure should be applied across nodes for
load balancing, only consider if the local partition has reached the threshold
limits instead of considering the size of the entire queue
This closes #3108.
Signed-off-by: Koji Kawamura <[email protected]>
> Load Balanced Connections can stop balancing across the cluster if all nodes
> report backpressure and all data destined for another node
> ---------------------------------------------------------------------------------------------------------------------------------------
>
> Key: NIFI-5745
> URL: https://issues.apache.org/jira/browse/NIFI-5745
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Affects Versions: 1.8.0
> Reporter: Mark Payne
> Assignee: Mark Payne
> Priority: Major
> Fix For: 1.9.0
>
>
> If a connection is load balanced, and backpressure is configured, we can
> possibly get into a situation where all nodes want to send data to other
> nodes, but those nodes are applying backpressure. If this backpressure is
> applied only due to flowfiles existing in the connection, waiting to be
> distributed, we can reach a live lock scenario.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
