[
https://issues.apache.org/jira/browse/ARTEMIS-3753?focusedWorklogId=761128&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-761128
]
ASF GitHub Bot logged work on ARTEMIS-3753:
-------------------------------------------
Author: ASF GitHub Bot
Created on: 22/Apr/22 21:53
Start Date: 22/Apr/22 21:53
Worklog Time Spent: 10m
Work Description: clebertsuconic commented on PR #4012:
URL:
https://github.com/apache/activemq-artemis/pull/4012#issuecomment-1106915511
note for myself (or you) so I won't forget...
- on the test, The other side of the mirror must have a cluster with Strict
load balancing... so messages will be load balanced when sent.
- on the implementation, I should add a routingType on RoutingContext where
I could set the routingType = OFF, and on PostOfficeImpl:;route (probably on
Bindings) I would look if the routingType on the context != null, I would
ignore any routingType and just the one from routing.
- The Mirror target will set routingContext.routingType=OFF, so messages
will always go to local queues on target.
- The test should be fixed with these options applied.
Issue Time Tracking
-------------------
Worklog Id: (was: 761128)
Time Spent: 3h 10m (was: 3h)
> Messages are not deleted in the mirror broker if they were distributed from
> the main to another broker in the cluster
> ---------------------------------------------------------------------------------------------------------------------
>
> Key: ARTEMIS-3753
> URL: https://issues.apache.org/jira/browse/ARTEMIS-3753
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Components: AMQP
> Affects Versions: 2.19.1, 2.21.0
> Environment: JDK 8
> JDK 11
> Windows 10
> Ubuntu 20.04
> Reporter: Mikhail Lukyanov
> Priority: Major
> Attachments: ArtemisClusterMirror.rar, brokerA1.xml, brokerA2.xml,
> brokerB1.xml, image-2022-04-01-18-04-36-472.png,
> image-2022-04-01-18-04-42-130.png, image-2022-04-01-18-05-12-973.png,
> image-2022-04-01-18-05-27-639.png
>
> Time Spent: 3h 10m
> Remaining Estimate: 0h
>
> When testing
> [Mirroring|https://activemq.apache.org/components/artemis/documentation/latest/amqp-broker-connections.html#mirroring],
> I found that it works unstable for brokers that are in a cluster. Messages
> are not deleted in the mirror broker if they were distributed ([_*initial
> distribution*_ and
> *_redistribution_*|https://activemq.apache.org/components/artemis/documentation/latest/clusters.html]
> too) from the main to another broker in the cluster.
> I have the following scheme, brokers A1 and A2 into clusters and A1 has
> mirror B1.
> *Steps*
> # Create 3 brokers (A1, A2, B1) with default user (root) and password (root)
> by _artemis create_
> # Replace their configuration with the configuration from the attachment
> (brokerA1.xml and etc.)
> # Run this brokers _artemis run_
> # Run ArtemisConsumer from maven project ArtemisClusterMirror.rar in
> attachment
> # Run ArtemisProducer from maven project ArtemisClusterMirror.rar in
> attachment
> # Wait for the producer to finish
> # Turn off consumer
> # Compare the number of messages queue Q1 in brokers A1 and B1
> *Expected Behavior:*
> The number of messages in the queue Q1 is the same (0), it is empty
> *Received Behavior:*
> The queue Q1 in mirror broker B1 contains 50 messages that were sent by the
> producer. There were no signals to delete these messages from broker A1 when
> moving them to another cluster broker A2 (with consumer). Screenshots in
> attachment.
> It is also not clear why the internal cluster queues with the _notif_ prefix
> are mirrored, in theory they should not. They also often contain messages
> that are not in the main broker.
> !image-2022-04-01-18-04-42-130.png!
> !image-2022-04-01-18-05-12-973.png!
> !image-2022-04-01-18-05-27-639.png!
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
