This role doesn't work for production deploys

[evrardjp]

I've noticed this isn't production ready deployment of graylog:

• There are assumptions in the role that software runs on localhost (I have PR to fix that for one part, but others like mongodb are still under the assumption of localhost)
• Even when ignoring the tests of localhost, the deploy doesn't work in multi-node. I've followed this: http://docs.graylog.org/en/2.4/pages/configuration/web_interface.html#using-a-layer-3-load-balancer-forwarding-tcp-ports . But the only way I can reach a page after authentication on the graylog web interface is to put 2 of my 3 servers into maintenance mode in my Load Balancer. I suspect that the authentication information are not shared between backends.

[evrardjp]

FYI I'm building a PoC integration of graylog for OpenStack clusters deployed with OpenStack-Ansible. Bridging communities would be great.

[evrardjp]

Is there someone here that can help at pointing me to the why of the authentication failure?

[mariussturm]

Sounds like your Graylog servers doesn't share one MongoDB instance. In that case they all run there own local database with own authentication data. Try to change the Mongo URL so that all servers point to one instance.

[evrardjp]

Thanks. So that's mongo DB the factor here. I will have a look at the role to know why the mongo DB isn't clustered by default when running the role on multiple nodes.

[mariussturm]

That's currently not implemented by this role (and actually not so easy to do).

[evrardjp]

I know the feeling :)

[malcyon]

I added a cluster example to the README. Hopefully, this suits people's needs. It is not perfect, but it does set up a MongoDB replica set, and gives a good idea of how to go about the whole thing.

[evrardjp]

I would give hugs on that last thing, but I know it's not covid safe. In the meantime, thanks @malcyon !

[malcyon]

My pleasure!