queue-master-locator option is deprecated and replaced by queue-leader-locator, which accepts different values.
https://github.com/rabbitmq/rabbitmq-server/issues/10702
According to this, if queue-master-locator is still used, “random” and “min-masters” will be mapped to “balanced” of queue-leader-locator.
https://www.rabbitmq.com/docs/clustering#replica-placement
However, in production, we find that the queue-master-locator option is not always correctly mapped, which can cause a node to abruptly terminate itself and subsequently lead to cluster partitioning.
“`
Nov 13 15:09:11 juju-d66aa4-2-lxd-16 systemd[1]: rabbitmq-server.service: Control process exited, code=exited, status=78/CONFIG
Nov 13 15:09:17 juju-d66aa4-2-lxd-16 systemd[1]: rabbitmq-server.service: Failed with result ‘exit-code’.
“`
The above is observed with RabbitMQ 3.9.13 on Ubuntu Jammy.
Since the queue-leader-locator defaults to “balanced”, which is overall a good default, there is no point in rendering the queue-master-locator for versions 3.9 and newer.
Also, the output of the `rabbitmqctl environment` command for versions 3.9 and newer will not show the “queue-master-locator” output by default.
Read more here: Source link
