[Gluster-users] glusterfs to replace nfs?

[Niels de Vos]

On Sun, Aug 17, 2014 at 09:59:02AM +0200, cvb at kruemel.org wrote:
> Hi,
> 
> maybe a daft, but hopefully easy question: I'm having some issues
> with nfs, statd, etc., which I'm unable to resolve since months (and
> which are on-topic elsewhere), and am therefore looking for
> alternatives to mount some resources on my storage server to a
> couple of VMs running here.

Some of the issues with NFS can occur with GlusterFS too, it just 
depends on what issues you're talking about. A pointer to the related 
discussions would be helpful.

> I made very good experience with glusterfs in the past, and hence
> tried it: Created (local-only) volumes with only one brick in them,
> and mounted these on the VMs.

Okay, so you are sharing a Gluster volume from a storage server to 
a virtual machine running on the storage server. This is a relatively 
simple configuration and should not be too troublesome.

> It appears to work - I'm just wondering if I'm naively doing
> something which might lead to issues...? And is there a (small)
> limit to the number of volumes I can create?

There is no hardcoded limit, but you should be aware that each brick you 
create starts its own glusterfsd process. So, in combination with each 
virtual-machine, you get a bunch of processes that may all compete for 
resources (RAM, CPU, bandwidth, ..). If the number of virtual machines 
and volumes grows, you may want to look into limiting resource usage 
with Cgroups or the like.

HTH,
Niels