New Server RAID

[karldenton]

Getting a new (2nd hand) server with 8 x 300GB SAS drives.

 

Is it best practice still to use 2 for OS and 6 in RAID 5 for data or put all 8 on RAID 5?

 

Even 6 giving 1.5Tb of data would be ok for now

[Joanne]

What about RAID 10? So 4 drives in RAID 5 which then mirror to the other 4 drives.

[SHimmer45]

[blue touch paper and stand well back]

RAID 5 isnt as popular for large arrays anymore rebuild times and risk of failure of another disk while its rebuilding.

10 is pretty much the go to for alot of things and would mean you wouldnt need to have 2 types of RAID in the same box if your looking to keep things simple.

 

last storage server i purchased through Dell had two smaller drives in a mirror for the OS and then the rest set aside for data.

[3s-gtech]

My usual with that range of disks would be RAID 1 on two drives for host OS/OS and the rest in RAID 10 for VMs/storage. I don't touch RIAD 5 now apart from on some of my NASs where it's okay for the job.

[Sagima]

All disks in raid 10 then partitioned is how I do mine.

[zag]

Having 8 x mechanical hard disks is going to be a very high risk of one of them failing!

[Achandler]

I often wonder what people do to mechanical disks, describing 8 disks as a "very high risk" of failure is completely misleading.

 

People ran mechanical disks for years and years and rarely had failures, some people haven't ever had any.

[Joanne]

I often wonder what people do to mechanical disks, describing 8 disks as a "very high risk" of failure is completely misleading.

 

People ran mechanical disks for years and years and rarely had failures, some people haven't ever had any.

 

The more you have the more chance you have of one failing. I have 6 in my server and have replaced about 8....

[aleach2]

I am in the same boat. About to start rebuilding a backup server which has had it's main array setup as RAID 0. I was wondering which would be best RAID 10 or 50? I will probably have 6 4TB mechanical disks.

 

On another point, I had a Hitachi SAN running a mixture of SAS and SATA drives and never had a failure in the 6 years of running it. As long as you buy good quality drives and keep then in a good environment then they should last.

[zag]

I often wonder what people do to mechanical disks, describing 8 disks as a "very high risk" of failure is completely misleading.

 

People ran mechanical disks for years and years and rarely had failures, some people haven't ever had any.

 

Yep mechanical disks are prone to failure, its something like 5% I believe. If you have 8 disks in your array then that's a 40% risk of failure as any one of the drives can fail.

 

To reduce the odds, reduce the number of disks!

 

NOTE: This is obviously subject to the disk manufacturer, heat and model type..

[3s-gtech]

I have some lovely HP SE326M1 servers with 20 or do disks in. Not been bad for disk failures considering the quantities, but obviously they have to be monitored!

[Joanne]

NOTE: This is obviously subject to the disk manufacturer, heat and model type..

 

Yeah don't run Seagates hot... that's my advice. Or even don't run seagates.

[SHimmer45]

people will still be running traditional disks in there servers while they continue to hit a good price point for capacity and performance.

i wouldnt have a problem choosing a HDD over an SSD if it met the requirements i was looking for.

i have seen mechanical drives in arrays which are 10 years old in heavily used file servers and they have been quite happy (temp controlled environment)

so the "unreliability" "chance of failure" on paper doesn't always translate to real world usage

 

im not very good at maths but a 40% chance of failure would seem a very high figure?

[Joanne]

im not very good at maths but a 40% chance of failure would seem a very high figure?

 

First glimpse I thought it was wrong, but each drive has a 5% chance of failure. With 8 drives you have a 40% chance of one of them failing (8 x 5).

 

But then it's a 0.025% chance of all of them failing (1 / (8x5) )

[SHimmer45]

i was looking at a calculator RAID Reliability Calculator - WintelGuy.com which made interesting reading, putting some average array sizes i couldnt get the annual drive failure rate to go above 10%.

 

but then again figures can be read in many different ways

[3s-gtech]

I'd say that sounds about right, IME the 40% would be over four to five years (the usual lifetime of the server).

[Cazale]

Raid 6 offers good protection against a second failure during rebuild (rebuild time is still a factor obviously, but weighed up against better efficiency in the use of overall drive space, once you get above 4 drives).

[CAWJames]

10 for us, pretty much the advice we get whenever raiding new or old hardware Just stay away from 5

[Cazale]

10 for us, pretty much the advice we get whenever raiding new or old hardware Just stay away from 5

 

Hardware vendors recommend using the configuration that uses the most drives? Who'd have thunk it!

[minimoo]

Dell recommend not to use Raid 5 and to use Raid 6 - on the basis that as drive sizes increased, they started seeing more problems with rebuilding raid 5 from hotspares i believe.