Hello, I have some questions about mirroring via AMQP broker connection and ways to adjust its configuration in case of bad network conditions.
For context, I'm building an Artemis topology across two datacenters. It is composed of two clusters (one on each datacenter) in replica mode (primary/backup pairs) relying on external quorum. Symmetric federation is enabled so that messages can be exchanged between both data centers. On top of that configuration, I'm now thinking about disaster recovery for the following requirements: Req 1 - Store a copy of messages in the other datacenter so that in case of a disaster no messages are lost. The Artemis sync mirroring feature covers this requirement as the client is blocked until the mirror ack the message. Req 2 - Reduced impact on message throughput in case of network instabilities: As clients are blocked with mirror sync, the message processing throughput is directly impacted by network instabilities (like latency issues), clients can be entirely blocked (and will timeout) if the target mirror broker is unreachable. As blocking a client for too long is not acceptable for my use case, I'm trying to understand if there is a way to adjust mirroring configuration when network instabilities arise. I have thought about a few options that could help to cope with network issues and that would allow client applications to continue using the broker at the cost of temporarily reducing message mirroring guarantees. Option 1 - Sync mirroring mode + switch to async when network is unstable It would be the best of both methods, although, messages would be temporarily at risk when in async mode. For now, I have found no way to switch from sync to async dynamically via the administration console, so I have some doubt it could be achieved. I'm not sure if it is something feasible via JMX calls or a custom plugin. A broker restart is possible to switch modes but it would close all client connections and that is not something possible in my case. Option 2 - Sync mirroring mode + stop the broker connection when network is unstable It allows client applications to use the broker but will stop message mirroring, messages are at risk until the broker connection is restarted. I have tried to test it and have stopped the broker connection via the admin console, but client applications are still stuck and cannot consume/produce messages. I don't know if it is normal behaviour or not, does someone have some insight about it ? Option 3 - Use async mirroring Not impacted by networking issues but at risk of losing some messages in case of disaster. I would be more than happy to know your opinion on the above options, if you have other ideas feel free to let me know. Many thanks J-P
