[
https://issues.apache.org/jira/browse/ARTEMIS-4558?focusedWorklogId=898963&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-898963
]
ASF GitHub Bot logged work on ARTEMIS-4558:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 10/Jan/24 12:37
Start Date: 10/Jan/24 12:37
Worklog Time Spent: 10m
Work Description: gemmellr commented on code in PR #4734:
URL: https://github.com/apache/activemq-artemis/pull/4734#discussion_r1447330865
##########
artemis-protocols/artemis-amqp-protocol/src/main/java/org/apache/activemq/artemis/protocol/amqp/connect/AMQPBrokerConnection.java:
##########
@@ -498,7 +498,7 @@ private Queue
installMirrorController(AMQPMirrorBrokerConnectionElement replicaC
Queue mirrorControlQueue =
server.locateQueue(getMirrorSNF(replicaConfig));
if (mirrorControlQueue == null) {
- mirrorControlQueue = server.createQueue(new
QueueConfiguration(getMirrorSNF(replicaConfig)).setAddress(getMirrorSNF(replicaConfig)).setRoutingType(RoutingType.ANYCAST).setDurable(replicaConfig.isDurable()).setInternal(true),
true);
+ mirrorControlQueue = server.createQueue(new
QueueConfiguration(getMirrorSNF(replicaConfig)).setAddress(getMirrorSNF(replicaConfig)).setRoutingType(RoutingType.ANYCAST).setDurable(replicaConfig.isDurable()).setInternal(false),
true);
Review Comment:
I have no concern around whether it can be managed or not. I'm saying that
my impression from the prior change that added the 'dual mirror' stuff and made
it 'internal' was that it was done more to have it treated a specific way (e.g
excluded) by other e.g mirror/cluster/etc related code in the broker and not
just to hide the queue from management. I am raising the question over whether
thats actually the case and if so whether any previously desired treatment
elsewhere wont occur as a result from simply marking it not-internal again.
This didnt occur to me before, but while typing that it did... I believe
Justin (or maybe you?) made a change to persist the 'internal' value at some
point...if so is the broker code actually going to end up dealing with a
mixture now where some are and some arent?
Issue Time Tracking
-------------------
Worklog Id: (was: 898963)
Time Spent: 1.5h (was: 1h 20m)
> AMQP Mirror ACKS should be idempotent
> -------------------------------------
>
> Key: ARTEMIS-4558
> URL: https://issues.apache.org/jira/browse/ARTEMIS-4558
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Reporter: Clebert Suconic
> Assignee: Clebert Suconic
> Priority: Major
> Fix For: 2.32.0
>
> Time Spent: 1.5h
> Remaining Estimate: 0h
>
> When I first developed Mirroring, I assumed sending the mirrored ACK on a
> aferACK and disconnected from any other transactions would be enough, with
> the caveat you could get a duplicate delivery on the target mirror in case of
> failures.
> I got some complains that this is not safe enough from some users, and I'm
> making this now idempotent.
> I took an overal mirroring hardening approach and I'm improving test coverage
> for this improvement.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
