[jira] [Commented] (ARTEMIS-3507) Smarter batching of acks

["Justin Bertram (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARTEMIS-3507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17424629#comment-17424629 ] 

Justin Bertram commented on ARTEMIS-3507:
-----------------------------------------

It's worth noting that batching acknowledgements is a performance optimization so it doesn't make sense to batch acks in a "low traffic scenario." This is similar to the {{consumerWindowSize}} setting which needs to be configured based on how quickly the client is expected to acknowledge messages (since you don't want to get into a starvation issue with unconsumed messages sitting in client buffers when other consumers are idle).

Is there a specific reason why you don't want to manually configure the {{ackBatchSize}} for your client?

> Smarter batching of acks
> ------------------------
>
>                 Key: ARTEMIS-3507
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-3507
>             Project: ActiveMQ Artemis
>          Issue Type: Wish
>          Components: Broker
>    Affects Versions: 2.17.0
>            Reporter: Lasse Lindgård
>            Priority: Major
>
> Using the default settings and the CORE api, I create a long running ClientSession, with a MessageHandler.
> Now when I consume messages I can see that the message acks are batched and delivered to the broker as configured in ackBatchSize. And the Message Count in the management console are only decreased when the acks are actually sent.
> In a low traffic scenario, with long running consumers, that might cause the acks "never" be sent or only to be sent on restarting the services.
> I wish that that could be improved somehow. And not by setting ackBatchSize=0. But in some clever way make sure that the acks are sent, even when the traffic is lower.



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