You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@nifi.apache.org by "ASF subversion and git services (Jira)" <ji...@apache.org> on 2022/04/04 18:30:00 UTC

[jira] [Commented] (NIFI-9847) Some processors may prevent NiFi from completing node offload

    [ https://issues.apache.org/jira/browse/NIFI-9847?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17517013#comment-17517013 ] 

ASF subversion and git services commented on NIFI-9847:
-------------------------------------------------------

Commit d8053601572d83609aaea10bcdbb092c4c262483 in nifi's branch refs/heads/support/nifi-1.16 from markap14
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=d805360157 ]

NIFI-9847: Switched LifecycleState to use a WeakHashMap to track Acti… (#5917)

* NIFI-9847: Switched LifecycleState to use a WeakHashMap to track ActiveProcessSessionFactory instances, instead of a regular Set that removed the instance after calling onTrigger. This was necessary for processors such as MergeRecord that may stash away an ActiveProcessSessionFactory for later use, as we need to be able to force rollback on processor termination

* NIFI-9847: Fixed checkstyle violation

> Some processors may prevent NiFi from completing node offload
> -------------------------------------------------------------
>
>                 Key: NIFI-9847
>                 URL: https://issues.apache.org/jira/browse/NIFI-9847
>             Project: Apache NiFi
>          Issue Type: Bug
>          Components: Core Framework
>            Reporter: Mark Payne
>            Assignee: Mark Payne
>            Priority: Major
>              Labels: listenhttp, load-balanced, load-balanced-connections, merge-content, merge-record, mergecontent, mergerecord, offload, processsessionfactory, stuck
>             Fix For: 1.17.0, 1.16.1
>
>          Time Spent: 0.5h
>  Remaining Estimate: 0h
>
> When a node is offloaded, NiFi stops and immediately terminates all Processors that are active in order to make their FlowFiles available for offloading.
> Some processors, however, hold onto a ProcessSessionFactory in order to allow for sessions to span multiple onTrigger invocations. Examples include MergeContent, MergeRecord, ListenHTTP.
> When these processors are used, and NiFi terminates the Processor, it is possible that some sessions will not be rolled back. As a result, the FlowFiles do not get offloaded, and NiFi will be stuck waiting for the offloading to complete.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)