clebert suconic commented on ARTEMIS-1328:

michaelandrepear: so just quickly why the increase of CHECK_QUEUE_SIZE_PERIOD
[2:26pm] michaelandrepear: from 100 to 1000
[2:26pm] michaelandrepear: what does that do?
[2:26pm] clebert: it will check if the queue can be back into direct deliver
[2:26pm] clebert: it’s meant for after low periods
[2:26pm] clebert: I think 100 milliseconds it’s too short
[2:27pm] clebert: it’s usually meant to after the consumer caught up
[2:27pm] michaelandrepear: ok cool
[2:27pm] clebert: so.. you have no messages pending
[2:27pm] clebert: beeing idle for a while
[2:27pm] clebert: you check again…
[2:27pm] clebert: checking every 100 Milliseconds was being too much I think
[2:27pm] clebert: maybe I should add that to the JIRA
[2:28pm] michaelandrepear: i just couldnt work out what exactly 100 to 1000 was 
specific if its just gut feeling thats cool
[2:28pm] clebert: do you mind if I copy & paste the chat here into there?
[2:28pm] michaelandrepear: i was thinking it was more specific reason
[2:28pm] clebert: gut feeling really
[2:28pm] michaelandrepear: of course not
[2:28pm] clebert: after running tests

> Delivery guard can take too long
> --------------------------------
>                 Key: ARTEMIS-1328
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-1328
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>    Affects Versions: 2.2.0
>            Reporter: clebert suconic
>            Assignee: clebert suconic
>             Fix For: 2.3.0
> There is a flush on delivery guard flush.
> if it's taking too long it will block the queue for a long period.
> This needs to be a quick check or keep doing async checks.

This message was sent by Atlassian JIRA

Reply via email to