Convy with devicemapper and also use Gluster?

Convoy
1

I have two identical servers.

We have spinning disks and a fast SSD. I have setup convoy to start with devicemapper on the fast SSD volumes. I followed the instructions on the github convoy page. So I have a convoy daemon running with devicemapper.

Now how do I use GlusterFS for the other spinning disk. It is a single volume. I guess I have two questions.

  1. When starting GlusterFS from the GUI, how do I specify where the storage “bricks” are used from? I looked in the YML files but did not see anything.

  2. If I start Gluster and then start Convoy Gluster which is a convoy daemon using the gluster volume. how does the convoy daemon on the host relate to this Convoy Gluster? I think they are not compatible.

Is it possible to accomplish what I want? Have two different storage pools? One shared across my machines (using gluster) and the other one being local (using devicemapper).

I see this on the Gluster Convoy page:

LIMITATIONS
Only one Convoy stack can be deployed once per one Rancher environment.

I see https://github.com/rancher/convoy/blob/c044d0073ecba40d50e8b94f32b6c04b9afe0d1e/docs/glusterfs.md the instructions on glusterfs. It sounds like I should start Gluster from the stacks and then add it to the Convoy daemon manually. So then my question is only #1, how do I specify the location of the gluster bricks.

Is it related to this? [Request] glusterfs should allow to choose the location of the "brick" · Issue #52 · rancher/catalog-dockerfiles · GitHub

2

Just a comment to this. I am thinking of just running Gluster on the hosts directly without docker. Then starting convoy with the driver and hosts.

I suppose my question was about the glusterfs catalog service and not about Convoy and Gluster directly.

I see there is a related question: Rancher catalog, GlusterFS, how to choose where bricks get created?