I know gluster 3.12.15 reached EOL long time ago. That doesn't meant that newer versions are better. I tested and run gluster 3.12.15 during years without a single glitch (well, with one or two, to be honest
). And it performs hugely better compared with newer versions, mostly because version 3 didn't have all those bells and whistles, but have no memory leaks, it faster and more stable than the newer versions, likely to advice to critic environments like hospitals. I love gluster, really, but devs are adding features in each version and they have a LOT of regressions and memory leaks because their roadmap is too quick and full of new features, at this pace, testing a version X.0 on production servers is a complete craziness, and the version becomes stable when it reaches the .8 onwards, .10 or .12 the better, but their release cycle should be more quiet and more focused in being a stable product. If you take a look on Red Had Gluster Storage, you'll see they are selling Gluster 3.4 (which must be 3.4 with all the patches until 3.12.15), and when the actual GlusterFS 6, 7 o 18 whatever version would be stable, Red Had will release Red Hat Storage X.
I must say on those critical scenarios I will stick with proxmox 5 and gluster 3.12.15 until I consider there is a rock solid version of gluster. Fortunately, those environments doesn't need to update their infrastructure often.
My problem is now with other scenarios, where huge MS SQL databases (+4TB) where working just fine with 3.12.15, but with 5.5, 5.8, 6.4 and 6.45downperforms... and turning back now is not an option. (I must admit 6.4+ seems to be quite fast compared with version 4 and 5)
Lets see if I found some gluster performance option to solve my problem