[
https://issues.apache.org/jira/browse/QPID-3422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14067446#comment-14067446
]
Fraser Adams commented on QPID-3422:
------------------------------------
I've had this conversation with Rob Godfrey in the past. I think that not
implementing ring buffers is the wrong choice. I appreciate that in many
circumstances you really don't want to be losing data but there are equally
valid circumstances where the laws of physics prevent more
transactional/handshake like behaviours from making any sense whatsoever.
The scenario that I care most about relates to high rate real-time producers. I
generally don't want to lose too much data, but it's stateless so not the end
of the world if I do. I've got a bunch of consumers, what happens if a consumer
fails? Well the data builds up on a queue, so far so good, but what happens if
it stays failed?
As I say I have a real time producer, so unless someone has recently
implemented a time machine I've got a problem! A ring queue lets the oldest
data fall on the floor. I can't really having this all flush to persistent
storage, TBH I care much more about ring queues than persistence with a
real-time producer.
This generally rules out the Java Broker for the scenarios that I care about.
> Implement ring queue on java broker
> -----------------------------------
>
> Key: QPID-3422
> URL: https://issues.apache.org/jira/browse/QPID-3422
> Project: Qpid
> Issue Type: Task
> Components: Java Broker
> Affects Versions: 0.10, 0.11, 0.12
> Reporter: Alex Rudyy
> Assignee: Alex Rudyy
> Priority: Minor
> Fix For: Future
>
> Attachments:
> 0001-QPID-3422-Implementation-of-ring-and-ring_strict-que.patch
>
>
> "ring" and "ring_strict" queue policies are not implemented on a Java Broker.
> Test AddressBasedDestinationTest.testBrowseMode fails against java broker
> because it uses a ring queue policy.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
