On Friday, February 10, 2023, 11:54:15 AM EST, Satish Patel <satish.txt@gmail.com> wrote:

Hi,

I went through that doc because it always shows up first in google search :)

My question was why the kolla default doesn't come with HA policies? I can surely add policy by myself and adjust requirements. I have also noticed without HA if any node fails then cause strange behavior in the entire cluster because of missing queues/data etc.

Personally I hate HA because the majority of messages are stateless. In my experience HA is a little stable when it comes down to node failure but with performance big no.

On Fri, Feb 10, 2023 at 11:15 AM Albert Braden <ozzzo@yahoo.com> wrote:

This has been discussed at great length; here are a couple of links that lead to interesting threads:

https://lists.openstack.org/pipermail/openstack-discuss/2021-November/026074.html
https://lists.openstack.org/pipermail/openstack-discuss/2020-August/016362.html

There used to be a useful document here:
https://wiki.openstack.org/wiki/Large_Scale_Configuration_Rabbit

That link now points to the “Large Scale SIG” home page. Does anyone know where the useful RMQ document is? I assume that it is buried somewhere under “Large Scale Journey.”

The bottom line is that, by default, kolla-ansible enables replication but not durable-queues, which is an unstable configuration. I believe that this may have been fixed in a newer version; we are on Train.

https://lists.openstack.org/pipermail/openstack-discuss/2020-August/016524.html

On Thursday, February 9, 2023, 10:48:11 PM EST, Satish Patel <satish.txt@gmail.com> wrote:

Folks,

I have deployed 3 node kolla controllers and notice rabbitMQ doesn't have any HA policy for queues. when i rebooted one of the nodes and found some messages got lost.

Any reason kolla doesn't use HA policies by default?

If i want to add policies then where should i specify them or any doc you can point me to define policies.