On 21 December 2017 at 00:19, Paul Tansom paul@aptanet.com wrote:
To emphasise the fact that RAID is not a backup I tend to use two situations I've been involved with. One is a Linux server that had software RAID on and had both drives fail together.
In the general case, having two independent pieces of hardware fail at the same time is very bad luck, and could happen whatever backup medium you choose.
But specifically for RAID: I always make sure I have different brands (or at the very least different models) of disk in my RAID arrays, so that a batch issue that causes them to fail at roughly the same age doesn't affect both simultaneously. For customers with preventative maintenance contracts we replace one of the two disks every year so that no disk is ever more than 2 years old.
However, both disks are by definition in the same location and connected to the same hardware. A power surge could take them both out together, as could fire or theft.
The flipside is that the most likely failure (one disk failing) is handled automatically and painlessly with little or no downtime, and the effort to rebuild is minimal. Of-course if you don't monitor the health of your array then the degraded array is no longer providing any redundancy but otherwise behaves as normal...
The other goes back further, to my NT sysadmin days. [...] Unfortunately the RAID hadn't finished recovering so the whole server, OS, data, the lot was corrupted. We sent the engineer away, got one of our IT staff on site and rebuilt the server from the backups - and never let a Compaq engineer near our kit again!
An external engineer will never be as careful with your data as you will!