Deletion of queue with unsynchronized followers

[Vishnu Vardhan]

Hi Team

Given that I have  "ha-sync-mode":"manual", "ha-promote-on-shutdown":"when-synced", and this in my RabbitMQ logs - "Mirrored queue 'statuschange' in vhost '/': Stopping all nodes on master shutdown since no synchronised mirror (replica) is available".

My mirrored queue is not functional anymore and I believe that I am facing the issue that is described here when there are leader shutdowns and no synchronized mirrors - https://www.rabbitmq.com/docs/3.13/ha#cluster-shutdown

In the above documentation link, it is mentioned to delete the queue and redeclare it to start using the queue again.

but in my case the queue delete operation is timing out(I have tried queue_delete using rabbitmqctl and also curl DELETE request to management API). In such cases is it safe to go ahead and use this command - rabbitmqctl eval 'rabbit_amqqueue:internal_delete({resource,<<"prod">>,queue,<<"x1">>}).' 

as mentioned in this ticket - https://github.com/rabbitmq/rabbitmq-server/issues/1333

and then expect that queue will be deleted and queue declare will work and the queue can be used.

Please clarify on the above query.

Thanks.

[Vishnu Vardhan]

Or is this command safe to use that removes queue records from mnesia tables

rabbitmqctl eval '
mnesia:dirty_delete(
  rabbit_queue,
  {resource, <<"/">>, queue, <<"my-queue">>}
).'

[jo...@cloudamqp.com]

Hi,

classic mirrored queues are not supported anymore. You can try to go with the first command (internal_delete) and then move on to dirty_delete. Eventually you need to move to quorum queues or non-mirrored classic queues. 

/Johan