I have faced that with quite old rabbitmq versions, like 3.7

As a workaround ha_queues worked nicely for me:

https://www.rabbitmq.com/ha.html#mirroring-arguments

вт, 18 окт. 2022 г., 15:45 Eugen Block <eblock@nde.ag>:

Are the remaining two nodes still member of a cluster? Can you share 
'rabbitmqctl cluster_status' from both nodes while the third is down?
How did you deploy openstack?

Zitat von Nguyễn Hữu Khôi <nguyenhuukhoinw@gmail.com>:

> Description
> ===========
> I set up 3 controllers and 3 compute nodes. My system cannot work well when
> 1 rabbit node in cluster rabbitmq is down, cannot launch instances. It
> stucked at scheduling.
>
> Steps to reproduce
> ===========
> Openstack nodes point rabbit://node1:5672,node2:5672,node3:5672//
> * Reboot 1 of 3 rabbitmq node.
> * Create instances then it stucked at scheduling.
>
> Workaround
> ===========
> Point to rabbitmq VIP address. But We cannot share the load with this
> solution. Please give me some suggestions. Thank you very much.
> I did google and enabled system log's debug but I still cannot understand
> why.
>
> Nguyen Huu Khoi