I want to balance disk space against fault tolerance. I would like fault tolerance to be able to handle multiple disk failures (as I may not be able to afford replacements disks or have the time for weeks possibly)
The main purpose of the home Linux server shall be a place to backup other machines too, store/share large amounts of data. So the data will be re-creatable most of the time. Storage of media (ie backup of my DVD’s, Cd’s etc)
I had a RAID10 array of 6 x 1.5TB but due to operator incompetence and laziness, I now have 6 empty disks 🙂 and a clean start.
One of the disks is definitely failing (over 55 error from smartctl and short and long test errors) so it will be sent away for warranty replacement – but I would still like to include it in the final array. Lets call the bad disk /dev/sdc
Machine has 6 sata ports and 2 IDE (with 2 CD drives). Dual Quad core Xeon, 16Gb RAM. And really 1 user most of the time.
[NB I may be able to remove a CD drive and add a 7th IDE disk just for the OS only to separate data/OS] Otherwise plan is to save 100GB partition and put the OS there (maybe mirror between disks)
RAID 6 sd[abdef], sdc as hot-spare (but gets sent for replacement soon) raid-devices=5 spare=1
RAID 6 sd[abdef], sdc as missing (but gets sent for replacement soon)
Option C) RAID 10 sd[abdef], sdc as hot-spare (but gets sent for replacement soon) raid-devices=5 spare=1
Option D) RAID 10 sd[abdef], sdc as missing (but gets sent for replacement soon) raid-devices=6 spare=0
Option A is seeming to be the best at the moment because I’ll get 4.5TB of space and room for 3 disk failures if I calculate it correctly.
This will all be done with mdadm soft raid.
Which do you recommend or are there better possibilities layouts that I could use?
Raid10 may not be able to handle two disks failing, so raid6 would be more reliable. It also gives more storage capacity. Option A only can handle the third failure if it happens after the rebuild onto the hot spare completes, but you only get 50% of the total space.
The odds of having 3 out of 6 disks fail are very slim, so I would be inclined to think that the loss of space ( and throughput ) is not worth having a hot spare. What you could do as a compromise is run without a hot spare, and in the event that you have a failure, and know you can not replace it for some time, and are worried that you might have two more failures, you could then reshape the array to a 5 disk raid6 and be back to being able to handle two more failures.
This requires that you use a filesystem that you can shrink since reshaping the array will reduce its capacity.
I recommend you check out Robin Harris (StorageMojo)’s blog Dear StorageMojo: cheap home bulk storage?. Robin, who writes about the data storage industry, thinks he should change his storage server “from RAID 5 to multiple RAID 0 stripes for speed and capacity.” But read the 50 comments for other opinions.
- ZFS to administer the file system and drives. But, “stay AWAY!! from ZFS unless you are a sysadmin type or SUN customer. I’m on zfs-discuss and the sheer number of problems people run into for simple stuff is mind-boggling”, says another commenter
- Drobo, but “too expensive per gigabyte”
- “mini-atx PC (the size of a breadbox) running either openfiler or freenas“
- Solaris 11 Express
- “RAID 0 is bad news and when it fails you will have major corruption issues.”
- ‘“The folks asking this question aren’t full time sysadmins..” Soon as I read this I thought “Get a NAS”.’, e.g. ‘QNAP 659 Pro+ with 6x 2 TB drives in a RAID-6 configuration.’
- “what is wrong with just high grade SATA spindles, an a mirror copy, same spec? Things like WE4 do not even power up unless needed…. And suddenly, for
I don’t claim this picks one of the options you put forward. But it is food for thought.
Me, I chose a NAS, with 2×1 TB drives in Raid-1 (mirrored). Every now again I rotate one of the drives to off-site storage and plug the previous off-site drive into the NAS.