How many drives do I need for RAID? [duplicate]

Question

I build a microserver to access my personal data and web services from my client machines. Since this is all the stuff I collected in the past 15-20 years, I'd like to keep it safe, so I'd like to use RAID this time. I have already checked that only 3.5" HDD-s have rational prices, so using 2.5" hdd or ssd is not an option. The motherboard supports RAID 0/1/5/10. I am currently selecting a case, but I am not sure whether it should have 3 or 4 drive bays for 3.5" hdd-s. The case I really like (Tt Core V21) has only 3 x 3.5" drive bays. The other case which is not that good if we are talking about cooling and air flow (BitFenix Phenom Micro-ATX) has 4 x 3.5" drive bays. What are the pros and cons using 3 or 4 disks if we are talking about RAID? (I guess it is something general, but if not, then I'd like to know it by each RAID version.)

Conclusions:

1. I think 3 drives with RAID-5 is sufficient for my needs.
2. I need a self-healing filesystem like ZFS or BTRFS (I chose the latter one), which can fix data degradation using RAID parity data. These filesytems can make incremental snapshots too, so they are great if you want a fast backup without halting the system. Using an event storage can do the same on application level, so that part does not really matter in my case.
3. People here never heard that backup does not protect against bit rot, but a self-healing filesystem with RAID most of the time does. Probably they are too focused on their RAID is not a backup mantra to learn new things...

Answer 1

Depends on what level RAID you want to do, and of course that depends on what you want to do with the RAID :)

Also don't forget that with Linux you can do software raid, no hardware support needed. What is nice about this is that the raid drives can be moved to another machine and be brought back up quick and easy.

Minimum of 2 drives for both RAID-0 and RAID-1. Minimum of 3 for RAID-5, 4 for RAID-10. Of course, you can always add more... if that is of interest, you may want to look into LVM and RAID+LVM....

RAID-0 stripes the data across both disks, which can help speed things up for doing work like video capture. Since both disks are evenly used for data, if you loose a disk the -0 in the name tells you how much data you'll be able to get back.

RAID-1 mirrors the disk, keeping the 2 drives in synch. If you have a failure of one disk, you can recover. If you have a hot spare defined (a 3rd disk) you may not even notice when it fails.

RAID-5 writes pairs of bits of data across 2 disks and writes a parity bit of the 2 bits it just wrote to a 3rd disk. And of course it shuffles which disk gets which bit. Loose one disk, you can recover. Again, if you have a hot spare defined (4th disk) you may not even notice.

RAID-10 is really RAID-1+0. Start by setting up 2 identical RAID-1 systems (4 drives), and then make a RAID-0 volume out of them. As long as one drive in each RAID-1 array is good, you can recover from a failure, and you get the speed boost of RAID-0. For seamless fail over you'll need a spare drive for each RAID-1 set up (2 more drives, 6 total).