Raid 5, Yay or Nay..

Si. · 18 Apr 2017 at 05:01

I'm currently running RAID-5 on my Synology NAS without any problems, I've been running RAID-5 on servers for 20 years and luckily never had a RAID failure (Other than those caused by human error). However I'm about to upgrade from my 4 bay NAS to an 8 Bay one and I'm starting to think I should now go to Raid-6.

What are peoples thoughts? I know Raid 6 will give me 2 disk redundancy, but is it worth the extra loss in space over adding in an additional RAID 5 disk? There are always worries about a corrupt RAID volume, and as drives get larger this risk increases, but if I'm currently running monthly RAID scrubbing on the array I'm hopefully mitigating this issue.

Any comments welcome.

BlizzardX · 18 Apr 2017 at 06:56

I personally use SHR with 2 disc redundancy. Above four data drives on Raid 5 I tend to move towards Raid 6 to give me confidence that I can withstand two drive failures... raid reconstruction is quite intensive and with more drives, there is a hgiher chance of a drive failing during the rebuild. If you have fast internet (100mbs+) then you might be able to mitigate the drive loss from backup and get the array up quickly.

Si. · 18 Apr 2017 at 07:24

Slow internet unfortunately.. I'm edging towards Raid-6 anyway so might go with that.

Armageus · 18 Apr 2017 at 07:35

What size drives and what is the NAS used for? What is your backup plan?

If it's data that is inconvenient but replaceable e.g. ripped blurays etc then by all means make use of the extra space in RAID5. Otherwise RAID6 is a better choice (and personally for anything super important RAID10 is even better due to less complexity and much faster rebuild times).

RAID Scrubbing is unlikely to be much benefit in RAID5, as all it is doing is comparing the data stripe to the parity data. If either is corrupt then all it knows is that one side is wrong, but which is side is correct? Not 100% sure on Synology's implementation but it is likely to just rebuild using the parity (even if that is corrupt). RAID6 should be slightly more reliable for scrubbing operations as you have 3 sources - (2 parity + data) so in theory it can look for 2 matching sources to correct the other.

Caged · 18 Apr 2017 at 08:01

RAID5 is most likely irrelevant at the sizes of disks and volumes that you'd encounter today.

BlizzardX · 18 Apr 2017 at 08:06

Synology Raid 6 implementation is using linux MDADM with either EXT4 or BTRFS as file systems. Scrubbing is comparing against parity but as there is no checksumming on write, bitrot can occur.

Xez · 18 Apr 2017 at 08:21

What size disks are you using?

Raid 5 is obviously brilliant for read speeds but like you say a raid failure whilst should be fine if you have large disks the rebuild time is terrible and during rebuild another disk can easily fail. Personally I would run raid 10 but a lot of disk usage is wasted.

SMN · 18 Apr 2017 at 10:18

Unless you have consecutive serial numbers, RAID5 is fine. Remember kids, RAID != Backup. You should have both.

Si. · 19 Apr 2017 at 03:19

Disks are 3tb.

Backups are ideal but I don't have the space anywhere.. None of the data is absolutely irreplaceable, but would be a royal pain to get back. Anything really important is also located on a local RAID 0 but that's nowhere near enough for everything.

nox_uk · 19 Apr 2017 at 11:15

backup disks are more important than raid for most people - if its jut say a film library (or other data where 90%+ does not change from day to day), then not watching a few films for a few days whilst you recover is a safer.

if it's mission critical, then raid becomes more important. re raid 5 vs 6 - how long does it take to purchase & rebuild a disk, and are you happy for to risk a disk not failing during that time?

nutcase · 19 Apr 2017 at 13:59

Si. said:
Disks are 3tb.

Backups are ideal but I don't have the space anywhere.. None of the data is absolutely irreplaceable, but would be a royal pain to get back. Anything really important is also located on a local RAID 0 but that's nowhere near enough for everything.

No way R5 then. R6 or preferably R10 if space/funds allow.

Caged · 19 Apr 2017 at 20:12

Also pray a lot

Si. · 20 Apr 2017 at 12:17

Raid-6 configured and parity check running

Little_Crow · 20 Apr 2017 at 13:37

nox_uk said:
how long does it take to purchase & rebuild a disk, and are you happy for to risk a disk not failing during that time?

The reasoning I've read around avoiding RAID5 with TB+ sized disks is not really around the risk of a further disk failing while you source a replacement. The biggest risk is the additional load you place all of the other disks in the array under during the rebuild - that is the time when your 2nd array-fatal disk failure is most likely to happen.

No RAID can be immune to this, but risk mitigation in a home setup can only go so far.

nox_uk · 20 Apr 2017 at 15:35

had never thought of that, but then that's never happened to me. run some weirdly clever hp raid level at work, ADM i think it is.

i like raid 1 & backups at home though

but then i'm not storing a stack load of data these days.

Armageus · 20 Apr 2017 at 15:44

nox_uk said:
had never thought of that, but then that's never happened to me. run some weirdly clever hp raid level at work, ADM i think it is.

ADM is Advanced Data Mirroring - nothing clever about it, just HP using their own name conventions for both RAID1 and RAID1+0 (and they call RAID6 "ADG" or Advanced Data Guarding)

Little_Crow · 20 Apr 2017 at 16:49

Never happened to me either, I run RAID5 at home but I'm fully aware of the risks associated.

You'll probably find the SAN is using SAS rather than SATA drives which gives improved error correction and better advanced warning of failure.
The advanced warning then triggering the use of a hot spare (Or releveling in the case of ye olde HP EVA SANs) potentially before the data on the disk is unreadable.

neil_g · 20 Apr 2017 at 21:54

Little_Crow said:
The reasoning I've read around avoiding RAID5 with TB+ sized disks is not really around the risk of a further disk failing while you source a replacement. The biggest risk is the additional load you place all of the other disks in the array under during the rebuild - that is the time when your 2nd array-fatal disk failure is most likely to happen.

No RAID can be immune to this, but risk mitigation in a home setup can only go so far.

That's why you always have a backup

nox_uk · 21 Apr 2017 at 08:01

Armageus said:
ADM is Advanced Data Mirroring - nothing clever about it, just HP using their own name conventions for both RAID1 and RAID1+0 (and they call RAID6 "ADG" or Advanced Data Guarding)

It's not just RAID 1 though is it? RAID1 = 1 redundant disk. We have it set up to have 2 redundant disks, and before you shout that's just raid 6, it's set in a pool of three disks. Over cautious? Maybe, but i also know how long it can take a approvers & purchasing departments to get their wotsits in gear

I also *think* it is clever enough to have improved read speeds as I seem to remember it reads 1/x of the data from each disk and combines them. Will check this though.

Hence me considering it clever compared to a normal raid setup.

Frozennova · 21 Apr 2017 at 12:22

nox_uk said:
It's not just RAID 1 though is it? RAID1 = 1 redundant disk. We have it set up to have 2 redundant disks, and before you shout that's just raid 6, it's set in a pool of three disks. Over cautious? Maybe, but i also know how long it can take a approvers & purchasing departments to get their wotsits in gear I also *think* it is clever enough to have improved read speeds as I seem to remember it reads 1/x of the data from each disk and combines them. Will check this though.

Hence me considering it clever compared to a normal raid setup.

Regardless of the number of disks, if the data is mirrored across the array it is still RAID1.