[ 
https://issues.apache.org/jira/browse/ARTEMIS-380?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Howard Nguyen updated ARTEMIS-380:
----------------------------------
    Attachment: Artemis General Practice.jpg

Hi,

Thank you for looking at the ticket.

By creating the maven project, I was trying to simulate the problem locally. 
Which I don't see it show up often.

I was just doing on our production artemis cluster very recently. I saw the 
problem again.

I have attached the diagram to describe the sequence of events.

A = active
B = backup
Host A = Host active
Host B = Host backup
Box = node

Column 1,2,3,4 from left to right is the sequence.

Column 1: Everything is fine. A is active, B is backup. Topology is checked.
Column 2: We shutdown Host A, Host B comes active. Host A stays down for 40 
minutes.
Column 3: We start Host A. It start as a backup doing it synchronized with Host 
B.
Column 4: Host A becomes active by "failback" mechanism defined in broker.xml. 
Host B becomes backup.

At column 4, I check topology on hawtio for host A
The Pair says a = B, b = A (for some cases) or B (for some other cases).




> Cluster Topology is wrong after update restart.
> -----------------------------------------------
>
>                 Key: ARTEMIS-380
>                 URL: https://issues.apache.org/jira/browse/ARTEMIS-380
>             Project: ActiveMQ Artemis
>          Issue Type: Bug
>    Affects Versions: 1.2.0
>            Reporter: Howard Nguyen
>            Assignee: Justin Bertram
>         Attachments: Artemis General Practice.jpg
>
>
> Setup 4 lives, 4 backups
> See configuration:
> https://github.com/khaing211/artemis-cluster-example/tree/master/src/main/resources/activemq
> Cluster Topology thinks
> backup is active, and live is backup for each pair, meanwhile it is not.
> After the following restarts:
> 1. all backups are restarted
> 2. all backups log say "backup announced"
> 3. all lives are restarted
> 4. all lives log says "server started with node id = xxxx"
> 5. Topology check through JMX / Jolokia say that the Pair[a = backup, b = 
> live] for each pair.
> This problem doesn't show up very often. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to