Morpheus multi sites deployment

[Ibrahim-GCC]

Hi Team,

Morpheus can be implemented across 2 datacenters. If so, what installation method did Morpheus recommend? If you could share some architecture for reference. It will help me understand much better.

[CuongDuong]

hello @greg , do we have Morpheus-Refercence-Archirecture-6.x now? If we have already, please share it with me. Thank you.

[cbunge]

To elaborate on Ollie’s point, Morpheus requires any MySQL 5.7 or 8.x technology. Our support troubleshoots environments our services deploy, and our team has since transitioned to using InnoDB clustering from Percona.

[#4]

Hello,

Morpheus does not recommend active/active multi-site configurations of the appliance controller due to replication and latency issues in Percona and RabbitMQ. Using passive/active sites works, but the customer must manage the replication and failover/failback tasks to ensure only one leader is active.

There could be some other guidance by the community where active/active multi-sites are reliable.

Greg Willis

[Ollie-Phillips]

I think this is the latest public resource, relating to Morpheus 6.3. No doubt 7 is in progress.

Note that for transactional DB we now deploy MySQL InnoDB clusters with MySQL Router, in place of Percona

support.morpheusdata.com

[Ibrahim-GCC]

Hi Greg,

Thanks for your valuable response. Will it be possible to get some reference link where we can find the active/passive model.

There could be some other guidance by the community where active/active multi-sites are reliable - This one I am not getting it could you please explain it bit more.

[#7]

This document is for 5.x; we are working on updating it for 6.x, which has a change for the RabbitMQ load balancer requirement. The information you are looking for is in the section labeled Disaster Recovery.

morpheusdata.com

Morpheus-Reference-Architecture-5.X-v3.1-2023-January.pdf

1340.36 KB

[cbunge]

Additionally clustering ElasticSearch across datacenters would require Enterprise ES licensing which is not included with Morpheus. You could optionally choose to lose that data which is mostly machine metrics that roll every 30 days by default.