[
https://issues.apache.org/jira/browse/ARTEMIS-5189?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17902338#comment-17902338
]
Justin Bertram commented on ARTEMIS-5189:
-----------------------------------------
You're seeing the expected behavior (as [~cathodion] mentioned). Even though
producers might send durable messages to a topic, the queues for non-durable
subscriptions are, by definition, non-durable. This means that none of the
messages for a non-durable subscription are stored on disk and therefore they
are not available on a backup.
To be clear, it's not that the backup broker "drops" or "discards" the messages
when it activates after the primary broker stops. The messages simply _aren't
there_ on the shared store when the backup activates.
Regarding performance, I recommend you actually try using durable
subscriptions. You might be surprised with the performance difference between
Classic and Artemis. Artemis performs significantly better than Classic in many
use-cases, especially at scale. If your consumers are keeping up with the rate
of message production I wouldn't expect much performance degradation with 300
queues & consumers on durable queues vs. non-durable.
> [Artemis Failover] [Topic] [Non-Durable Consumer] Lost messages during
> failover switch
> --------------------------------------------------------------------------------------
>
> Key: ARTEMIS-5189
> URL: https://issues.apache.org/jira/browse/ARTEMIS-5189
> Project: ActiveMQ Artemis
> Issue Type: Bug
> Affects Versions: 2.38.0
> Reporter: Liviu Citu
> Priority: Critical
> Attachments: broker_backup.xml, broker_primary.xml
>
>
> I am currently migrating from Classic Broker to Artemis Broker and I also
> want to enable the Failover (Shared-Stored) switch. I have configured Artemis
> with Failover (Shared-Store) and during the failover switch, topic consumers
> loose events before reconnecting. When using durable consumers this is not
> happening so the problem (as expected) is only replicable for non-durable
> consumers.
> The reason for which I want to avoid using durable consumer is the
> performance impact on the broker side when there is a big number of messages
> and consumers for them. For a live client using Classic Broker I have
> observed a big impact and slowness of the broker under high load (thousands
> of messages with more than 300 durable consumers). Indeed I am not able to
> test the same with Artemis broker as I do not have access to that environment
> however my expectation is that the Artemis broker will behave similarly.
> Is there a parameter in Artemis broker to make it wait a bit after the switch
> before dropping topic message, so that previous active consumers to have
> time to reconnect and consume the messages?
> Attached the Artemis Brokers Primary/Backup configuration.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
For further information, visit: https://activemq.apache.org/contact
