• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

3PAR Raid Sets

bigdogchris

bigdogchris

Fully [H]
Joined
Feb 19, 2008
Messages
19,076
My sales guy wants me to configure SAS CPG RAID 5 as 4+1 and NL CPG RAID6 as 8+2 but the HP 3PAR best practices guide said RAID 5 should be 3+1 and RAID6 6+2.

We'll be gaining capacity but losing fault tolerance. What do you guys think?
 
Last edited:
If your workload fits within the abilities of R6 speeds there is no reason not to go double parity. If you can increase your uptime during failure it is worth it, especially with multi-terabyte drives! In the same number of drives, R6 beats R5 plus hot spare because it saves you in case of a second failure during rebuild.
 
ive never understood the resoning behind 2 hotspares on raid 6 arrays. theres already double parity, and if your not monitoring it close enough to notice when the 1 hotspare has taken over you have other issues that raid will not fix.

i also try to never make a raid array with the minimum allowable number of drives. with all the aligning and overhead loss you actually loose more than is gained back.

definitely try to go all raid 6, and ONLY use more than 1 hotspare if you have a full chassis with lots of arrays and the hotspares are global. up to 12 drives in any number of arrays in the chassis i normally only use 1 global hotspare. (as long as drive age is all similar)
 
I would say the same. Go with R6. There really isn't a need for the hotspare unless you are dealing with truly mammoth arrays which really need a global spare. R5 at this point is reserved for chassis constrained setups.
 
DISCLOSURE: I work for HP Storage.

One point that I think is missing with the hot spare conversation is how HP 3PAR uses spares. There isn't a drive that is flagged as spare and sits there idle waiting for a failure. Because 3PAR uses fine grain virtualization to spread data across all drives in the array, we don't use spare drives but rather spare capacity that is spread across all the drives.

In addition, with the 3PAR ASIC, rebuilds (among lots of other things) get an assist so they do happen faster than a typical array.

I would recommend following the best practices document. Another suggestion is that there is a dedicated 3PAR users group with lots of 3PAR users with lots of experience. You might ask there as well. And note it isn't an HP users group but was started by a customer.

Here's the most recent HP 3PAR best practices guide (May 2015).

Happy to answer any questions.
 
Zedicus said:
ive never understood the resoning behind 2 hotspares on raid 6 arrays. theres already double parity, and if your not monitoring it close enough to notice when the 1 hotspare has taken over you have other issues that raid will not fix.

i also try to never make a raid array with the minimum allowable number of drives. with all the aligning and overhead loss you actually loose more than is gained back.

definitely try to go all raid 6, and ONLY use more than 1 hotspare if you have a full chassis with lots of arrays and the hotspares are global. up to 12 drives in any number of arrays in the chassis i normally only use 1 global hotspare. (as long as drive age is all similar)
Click to expand...

RAID 6+2 doesn't mean two spares - it means 6 data and 2 parity drives.

3PAR is very different from a traditional RAID so the normal "what I've done with traditional RAID" could lead you down the wrong path.
 
Sorry for the multiple "answers" to this. I asked one of my storage architects to help me give a better answer (beyond what is in the best practices guide). Here's what I got:

The rule of thumb for best practice is to configure things so that the number of enclosures behind a node pair can be divided evenly by the RAID set size for RAID5 or RAID1, or half the set size for RAID6. You confused by that - read on.

That is because if the system allows for it, we provide an additional level of availability by creating the LDs for a CPG by taking one chunklet from each of the drive enclosures in order to create a RAID set. Each LD can have multiples of these RAID sets and we widely stripe among them. (I’m simplifying here.)

So for example, if you have a 2 node system with FC drives in each of 4 enclosures, then R5, 3+1 works best. But if you have 5 enclosures, then 4+1 is best. If you have 8 enclosures, then either 3+1 or 7+1 is good. For RAID 6, we can provide the same availability by taking 2 chunklets from each enclosure. So 6+2 works best with 4 enclosures, 8+2 with 5 enclosures, and so on.

However RAID sets cannot span node pairs. So if you have 4 nodes, then you would need 8 total enclosures for 3+1, 10 enclosures for 4+1, etc.

The advantage of this is that if we lose an entire enclosure, then we still have access to all data because we can rebuild it from the remaining data and parity. This is only best practice; we can still create other RAID set sizes but they may not have the same level of availability. The guidelines can be found in the config guides for each product.
 
Calvin,

Gotcha. Dividing evenly like that allows even distribution of cage level striping for better fault tolerance.

Another question ... I have some physical disk showing fully provisioned space, yet every VM so far has been set as thick eager zero and has had free space zeroed. Is there a way to tell what volumes specifically have fully provisioned content? I can't seem to find that at the volume level, only disk tier level.
 
You must log in or register to reply here.
Back
Top