high-availability tests

[LeonardoSalvucci]

Hi, first of all congrats for this very good work. I'm working a lot with clusterized rabbitmq and this is a good start point for make a solid clusterized conteinerized rabbitmq service.

Did you do some tests with nodes (containers) down when publishing and how cluster reacts. I mean... how much time takes the master to promote, if some messages are lost and things like that?

Perhaps we can design some tests and share results.

Thanks again!

[bijukunjummen]

Hi @LeonardoSalvucci, thanks for your kind words. I did this purely as a learning exercise however and don't have any production clusters running based on this code, however if you have any tests I would be more than happy to collaborate and validate them.

[ypereirareis]

Hi @LeonardoSalvucci (hello @bijukunjummen )

I built a docker based project that allows to run a highly available RabbitMQ cluster.

You can test/benchmark many scenarios:

• Node failures
• Network partitions
• Exchanges, Queues, Messages persistence and mirroring
• ...

You can run parallel consumers and producers against HAProxy. It load balance every request on the nodes.

Have a look at https://github.com/ypereirareis/docker-rabbitmq-ha-cluster

Bye !