Backups are a tricky thing as everyone has different levels of tolerance to risk as well.
For starters, no RAID is not a backup. RAID should be used to increase performance and/or resilience and/or to maintain availability but should never be considered a backup.
Secondly there is also a difference between having a backup and disaster recovery!
Also always remember a backup is only any good if you can restore it, and KNOW that you can, we’ve all heard the story of the guy diligently backed up his data, then when he came to needing it found it was all corrupt and unusable.
So I’ll now spew forth some thoughts, these are based on MY approach to data risk, you maybe more or less paranoid than me!
I would never rely on a backup held in the same physical machine, I do often have backups/data duplicate in the same machine and make use of them but it would never be my only backup, you’re one power surge or blown PSU away from losing it all. Also since they’re in the same machine any silent data corruption due to bad hardware could nuke your backup as well as your primary.
I also would never rely on only having a single backup. If something bad happens to your primary you are now completely reliant on your backup, both being good and accessible, and also NOT failing during any restoration attempt, although unlikely it is a risk.
My setup is as follows:
all my machines have primary storage, sometimes this is RAID, sometimes not depending on my requirements in that machine.
Some of my machines also have secondary storage which contains data duplicates of important working data, this is regularly kept updated with rSync and is used purely to keep me running in a case of primary storage failure or a case of the whoopsies I’m deleting or breaking something.
I also have a primary NAS located on site which uses ZFS with a bunch of mirrored vdevs underneath. I use ZFS as it’s highly resilient, scalable and is self healing as it can identify data corruption AND fix it. It also has excellent replication and snapshot capabilities. I’ve been using it since it first appeared and it’s a VERY good filesystem but takes a little time to get the best out of it if you’ve not come across something like it before.
All my machines backup to this NAS and it also holds shared data used by multiple machines and is storage for my VM lab.
I also have a secondary ZFS NAS similar to the first on site, data is replicated from the primary using ZFS replication so this NAS is always ready to step in for the primary if that has issues. I do periodically failover to it just to test.
I also have a smaller capacity ZFS NAS that is kept offsite at a friends house, critical data (ie: personal files and important work, but not transient data) from the primary NAS is replicated here daily so this is my first port of call should anything drastic happen to my place! He has a reciprocal arrangement with me and I host his backup NAS.
As another layer of protection I also have an account with rSync.net and I replicate critical data to them nightly, so this is my next port of call if needed.
* in my primary NAS I use two disk controllers and each mirror pair has one drive on each controller, and I use different manufacturers drives for each half of the mirror. This helps mitigate against both bad drive models and controller issues.
For secondary and offsite NAS i forgo the extra controller but still mix disk brands.
The offsite NAS is smaller capacity as it only stores critical data but it’s setup as a quad mirror as I rarely see it physically so this gives added redundancy and also more leeway in case I do have a disk failure I don’t have to be quite as quick to sort it.
some may call my setup overkill but I’ve not yet lost any important data in the last 20 years and my current setup is more reliable and resilient than it was 10-15 years ago.
