This was recently posted on comp.sys.raspbery-pi:
The current [failed] disc is a 2TB, largely because it was what was available at the time. Most of the time it is less 2% used, so it is rather bigger than it needs to be, but is being used on a continuous basis.
So what is a suitable replacement HDD or SSD ?
Here's my, somewhat edited, answer:
I recently replaced a 37000 hour 120GB Hitachi TravelStar HDD in my ten year old Lenovo R61i laptop when smartd showed that seek errors were starting to occur and the throughput rate was dropping. Since the machine's SATA interface can't handle disks bigger than 200GB and hard disks this small are no longer available, the disk got replaced with a 128GB Sandisk SSD. However, because smartd says the SSD has only had 2 hours use so far I obviously can't say anything useful about its durability. What I can say is that it has made long, disk intensive tasks, such as booting the system, much faster while having little effect on more normal tasks. I put this lack of effect on normal activities down to the way Linux uses RAM as a large disk cache: this laptop runs Fedora 25.
At almost the same time a 49700 hour 3.5" 250GB Hitachi Deskstar in a dual-Athlon desktop system started showing similar problems and was replaced with a 500GB WD Blue about 250 hours ago, so again I can't yet say anything useful about its durability.
One observation is that the TravelStar remained readable and bootable throughout the replacement process, though it became subjectively quite slow, so I was able to copy its contents to the SDD by using CloneZilla booted off a DVD. However, the same cannot be said of the DeskStar, which deteriorated very rapidly. Its partition table became corrupted, which made the disk unbootable and also prevented CloneZilla and dd from copying anything from it to the new disk. Consequently, I had to do a clean Fedora 25 install from a live DVD and retrieve the contents of /home, /usr/java and /usr/local from the backup I'd made immediately before starting to replace the disk. The disk swap has made the machine a little faster because the Deskstar was an unbuffered drive while the WD Blue has a built-in 64MB buffer. Like the laptop, this has made disk intensive activities like booting a bit faster but has had little visible effect on more normal activities.
If I'd had my brain engaged I would have replaced the dead 3.5" Hitachi TravelStar with a 500GB or 1 TB WD Red because this system is run 24x7, the MTBF and warranty period for the Red series are both much longer than for the Blue and the annoyance factor of replacing a disk increases at least linearly with disk size.
Going beyond that, at some point the annoyance factor gets big enough to justify replacing a single HDD with a RAID farm so you can hot-swap failing drives.
Having an offline backup is still, IMHO, mandatory if you go that way because online drives are still subject to destruction from mains strikes, house fires and burglary.
Hearsay and internet sources say that a consumer-grade SSD can fail completely with little or no warning while an enterprise SSD tends to fall back to read-only at EOL so you can at least pull the data off it when it fails. Whether this (and cost) influences your choices depends a lot of how much you value the data and how frequently you make offline backups.