[jira] [Commented] (NIFI-8039) Decrease the initial resource reservation of Listen processor family

["ASF subversion and git services (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/NIFI-8039?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17240588#comment-17240588 ] 

ASF subversion and git services commented on NIFI-8039:
-------------------------------------------------------

Commit 39f8a008d4bde2024eb804ff7017542a6e86b572 in nifi's branch refs/heads/main from Bence Simon
[ https://gitbox.apache.org/repos/asf?p=nifi.git;h=39f8a00 ]

NIFI-8039 Adding properties to ListenTCP in order to allow refine behaviour under higher load; Refining thread pool for better scalability

NIFI-8039 Review findings; refining thread pool to be able to scale down properly when not under load
NIFI-8039 Answers to PR comments

This closes #4689.

Signed-off-by: Peter Turcsanyi <tu...@apache.org>


> Decrease the initial resource reservation of Listen processor family
> --------------------------------------------------------------------
>
>                 Key: NIFI-8039
>                 URL: https://issues.apache.org/jira/browse/NIFI-8039
>             Project: Apache NiFi
>          Issue Type: Bug
>          Components: Extensions
>            Reporter: Simon Bence
>            Assignee: Simon Bence
>            Priority: Major
>             Fix For: 1.13.0
>
>          Time Spent: 1h 10m
>  Remaining Estimate: 0h
>
> ListenTCP processor and other processors in the listen family are reserve resources at initialisation based on properties on the processor. This includes:
>  * Executor in SocketChannelDispatcher is created using fixed thread pool
>  * Byte buffer pool is build up based on Max Number of TCP Connections which with high connection can consume high amount of memory
> In some use cases this is undesirable and might even cause failure of the initialisation. In order to avoid this I suggest to remove the buffer pool and replace the fixed thread pool into a more flexible (scaling) implementation.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)