I'm doing web & mail hosting for myself and several small structures like our
local school and a handful of local companies. Up until recently these hostings
have happened on "bare metal" root servers using CentOS 7. One main server is
hosting most of the stuff: WordPress sites, one OwnCloud instance, Dolibarr
management software, GEPI learning platform, Postfix/Dovecot mail server,
Roundcube webmail, etc.
This setup has become increasingly problematic to manage, since applications
have more and more specific requirements, like different versions of PHP and
corresponding modules.
So I decided to split everything up nicely into a series of virtual machines,
each one with a nicely tailored setup.
I have a couple of sandbox servers, one public and one local, running Oracle
Linux 7 (a RHEL clone like CentOS). I played around with it, and KVM-based
virtualization already works quite nicely.
While looking for documentation, I stumbled over oVirt, which I didn't even
know existed until last week. Before I dive head first into it, I'd be curious
to know a few general things.
1. Would it be overkill for a small structure like mine?
We’ve seen people with 70+ machines being managed by shell scripts and libvirt, and also oVirt users with just a single host. Those would be extremes. I would personally start using oVirt once I have more than 4-5 hosts or so, and/or more than 20 VMs.
2. Will I be able to do HA on a series of modest KVM-capable root servers even
if they are located in different datacenters across different countries?
I think it mostly depends on the network latency and speed. Having HA with servers in different countries may be problematic.
yeah, in a stretched cluster setup it would work, but the expectations are that the link is almost as local. It does work ok with several hundred ms latency. but it needs to be reliable.
3. One problem I couldn't resolve using a bone-headed keep-it-simple KVM setup
is backup. For my bare-metal servers I've been using incremental backups using
Rsnapshot for years. Here's a blog article I wrote on the subject:
https://blog.microlinux.fr/rsnapshot-centos-7/
Unfortunately I can't use this approach with huge QCOW images, at least not
without jumping through burning loops.
Is there an easy way to perform remote incremental backups with oVirt?
There are a few backup vendors that provide solutions for backing up oVirt VMs, you can just query google for them.
I think none of the vendors are ready with their products to use incremental backup just yet. it’s close though.
you can also just adapt and use examples[1] of the incremental backup feature[2]. if you’d want to just hack up some DIY backup solution.
incremental backup us a fairly recent addition to oVirt (and to qemu/libvirt in general)
