[
https://issues.apache.org/jira/browse/NIFI-2836?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15552131#comment-15552131
]
ASF GitHub Bot commented on NIFI-2836:
--------------------------------------
Github user mcgilman commented on the issue:
https://github.com/apache/nifi/pull/1109
Looks like lots of debug level logging was added around the wait/notify of
the monitor used to lock during requests. However, in the case there was a
dissenting node in phase one of a two phase mutable request there is not
logging around the notify.
Additionally, it looks like when any node fails the onCompletion callback
is invoked immediately which triggers the notify signaling all requests have
completed when that is not the case.
> Cluster - Request Replication
> -----------------------------
>
> Key: NIFI-2836
> URL: https://issues.apache.org/jira/browse/NIFI-2836
> Project: Apache NiFi
> Issue Type: Sub-task
> Components: Core Framework
> Reporter: Matt Gilman
> Assignee: Mark Payne
> Priority: Critical
> Fix For: 1.1.0
>
>
> When a request is replicated throughout the cluster, the cluster coordinator
> needs to ensure that request completes (including both parts of the two-phase
> commit if applicable) prior to replicating the next request.
> Looks like a write lock is obtained while submitted a background task. The
> actually work completes outside the locks scope.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
