Hacker News new | past | comments | ask | show | jobs | submit login

I've been using Gluster in a production environment for over a year now with an absolutely perfect track record. I use it for the "centralized" filesystem for a virtualization cluster (using KVM). Currently, I have three virt hosts setup with dual NICs each to get 2Gbps throughput, sharing a RAID1-style filesystem (with priorities to the local copy for r/w speed).

I was (and still am) fairly skeptical about using Gluster in this role for a production environment having come from a fiber/SAN shop before. However, I gained a lot of faith in Gluster after a non-incident a few months back. One of our core switches fell offline (the one all three virt hosts were plugged into) and remained offline for about an hour I believe. When the switch was brought back online, Gluster managed to put everything back together and continue operation without any human intervention. All 20 or so VMs were still humming along like nothing had ever happened. fsck confirmed this.

Even in the case of a failure (which I've not experienced), your data is still sitting on disk (at least that's how it is with RAID1, I don't know how it's RAID0 setup works) -- you'll always have access to the raw files Gluster uses.




Glad to hear its being used successfully. I'm betting things improved greatly in the last couple of years. Thanks for sharing this. :)


It definitely has. Two years ago, I even had issues with some features of KVM/QEMU (live migration, for instance), let alone trying to run that on a distributed FS. It's amazing how fast these techs are coming along.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: