You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@activemq.apache.org by "Matt Pavlovich (Jira)" <ji...@apache.org> on 2021/01/15 18:13:00 UTC

[jira] [Updated] (AMQ-3226) persistent messages expire without being saved to ActiveMQ.DLQ if a master-to-slave failover took place - issue with JDBC master-Slave cluster

     [ https://issues.apache.org/jira/browse/AMQ-3226?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Matt Pavlovich updated AMQ-3226:
--------------------------------
    Labels: close-pending  (was: )

> persistent messages expire without being saved to ActiveMQ.DLQ if a master-to-slave failover took place - issue with JDBC master-Slave cluster
> ----------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: AMQ-3226
>                 URL: https://issues.apache.org/jira/browse/AMQ-3226
>             Project: ActiveMQ
>          Issue Type: Bug
>          Components: Broker
>    Affects Versions: 5.4.2
>         Environment: Sun Solaris + JDK 1.6
> a JDBC cluster of 2 nodes running apache-activemq-5.4.2-fuse-01-00
> using MS SQL Server 2008 as a database for JDBC
>            Reporter: Oleg Kiorsak
>            Priority: Critical
>              Labels: close-pending
>
> Setup : 
> Sun Solaris + JDK 1.6
> a JDBC cluster of 2 nodes running apache-activemq-5.4.2-fuse-01-00
> using MS SQL Server 2008 as a database for JDBC
> 1. 1 client (NMS STOMP , if that is any relevant) sends 10 persistent messages with TTL set to 5 mins
> 2. after all messages have been sent and seen in the destination queue (via SQL query)
> but well before the time message are about to even start expirying
> do a 'kill -KILL' of the activemq process on node1 ("master") - so 'failover' take place 
> - node2 ("slave") now becomes new "master" and resumes connections/etc
> 3. in 5 mins time all 10 messages sitting in the destination queue do expire, 
> but they do NOT end up in ActiveMQ.DLQ as they should...
> this is really bad - 
> basically the failover event breaks the promise of "guaranteed" delivery of persistent messages



--
This message was sent by Atlassian Jira
(v8.3.4#803005)