On Mon, Aug 01, 2016 at 07:24:39PM -0500, Nathan Neulinger wrote:
Essentially, gluster is a cluster distributed filesystem. It uses local storage on each 'server' node, and that can then be mounted on clients. It's intended for high performance compute resources, but works quite nicely for freeswitch backend.
In my case, I have 4 servers accessing the cluster - 2 freeswitch nodes that are setup as an HA pair using keepalived (article already in wiki for that), 2 web server front end nodes for a local custom UI.
On the backend, I have the three gluster nodes for file service (odd number to insure quorum) and 3 Percona XTraDB mysql nodes also operating as a cluster.
Everything except the two freeswitch instances is running as a VM. The FS nodes are running on real hardware.
The key difference with NFS is that the filesystem itself is redundant - so you can take any one of those backend nodes out of service and it isn't visible (except as a momentary hang) to the servers that have the filesystem mounted.
I appreciate all the info. Gluster sounds great and seems like it works out well in a situation like this. Did you try a document storage DB before going to gluster, or was gluster already what you were using so it just fell into place?
-- Nathan