I am currently planning a decent overhaul of the hardware in our co-lo cab. Right now, the plan is migrating the existing xcp-ng+xo ecosystem to 3x Dell R440s as compute nodes tied to a TrueNAS X10 as storage node. Ultimately, with HA, DR, and all that, but I can worry about that later.
What I’m trying to wrap my brain around right now is a somewhat optimal starting point with regards to snapshots and backups. I already currently have XO running weekly full and nightly delta backups, and will likely continue that, if not also modifying the schedule.
When would I want to use snapshots within XO vs TrueNAS? I know there’s always the whole “it depends” thing. After watching Tom’s latest NFS vs iSCSI video, my instinct is to create the initial VM SR with NFS, mainly because of thin provisioning, and because I don’t think any lack of speed compared to iSCSI would affect us much in our enviroment. Long-term, I can always plan out any iSCSI volumes as I need them. Then, I am thinking using XO for snapshots, within the same SR, and backups to another dataset. TrueNAS can then replicate those backups offsite. For any shares/datasets not directly being used by XO, I am thinking using TrueNAS for snapshots/backups with those.
I really am quite pleased with xcp-ng+xo, as it just seems to work… unlike my experience fighting with ESXi 6.7. When you dig into the cli of ESXi, it seems more and more like it was just bolted together and there are layers of legacy systems that don’t always play well together. Plus, the incompatibilities with 2018 firmware (because that’s all you could get without paying a bunch for what should be free) on EOL hardware that do not show up for a few months, and then you’re scrambling trying to figure out how to get the VMs out of that environment. Ugh.
Anyway, any help would be appreciated! Thanks!