VMware Cloud Community
b1izzard
Contributor
Contributor
Jump to solution

vSAN local RAID

Noob here.  I somewhat understand how vSAN works in a cluster, but how does it work locally?  Let's say I have mirroring setup and it is writing the same data to both nodes.  At the node level, I have 5 capacity disks in my disk group.  Does it just stripe as a raid 0 across all the capacity drives?  In the raid 0 scenario, I would imagine 1 drive failure would take down the entire node.  If that is the case can you do raid 5 or 6 locally or is that a bad idea and not supported?

Reply
0 Kudos
1 Solution

Accepted Solutions
TheBobkin
Champion
Champion
Jump to solution

Hello b1izzard​,

If for instance you have a RAID1 100GB vmdk, it will have a 100GB replica each residing on a single Capacity-tier disk on each of the nodes - if you lose that particular disk due to a failure, the now missing replica is rebuilt from the other copy onto another available disk on the node that lost a disk - losing a single disk doesn't fail the entire Disk-Group unless it is a Cache-tier device or it is a deduplication-enabled cluster.

I mention '100GB' vmdk above as once Objects get larger than ~250GB (or when using a Storage Policy with Stripe-Width >1) it starts RAID0 striping components of each replica - however if this occurs due to size and not Storage Policy, it may and can place multiple stripes on the same disk, whereas with Storage Policy with Stripe-Width >1 it will specifically aim to stripe it across as many disks (on each host) as the rule in the Storage Policy states.

"If that is the case can you do raid 5 or 6 locally or is that a bad idea and not supported?"

Nope, there is no way to RAID5 components across disks on just one node as this is fairly pointless when we are mainly relying on (and designing for) node/cluster level redundancy.

Bob

View solution in original post

Reply
0 Kudos
3 Replies
TheBobkin
Champion
Champion
Jump to solution

Hello b1izzard​,

If for instance you have a RAID1 100GB vmdk, it will have a 100GB replica each residing on a single Capacity-tier disk on each of the nodes - if you lose that particular disk due to a failure, the now missing replica is rebuilt from the other copy onto another available disk on the node that lost a disk - losing a single disk doesn't fail the entire Disk-Group unless it is a Cache-tier device or it is a deduplication-enabled cluster.

I mention '100GB' vmdk above as once Objects get larger than ~250GB (or when using a Storage Policy with Stripe-Width >1) it starts RAID0 striping components of each replica - however if this occurs due to size and not Storage Policy, it may and can place multiple stripes on the same disk, whereas with Storage Policy with Stripe-Width >1 it will specifically aim to stripe it across as many disks (on each host) as the rule in the Storage Policy states.

"If that is the case can you do raid 5 or 6 locally or is that a bad idea and not supported?"

Nope, there is no way to RAID5 components across disks on just one node as this is fairly pointless when we are mainly relying on (and designing for) node/cluster level redundancy.

Bob

Reply
0 Kudos
IRIX201110141
Champion
Champion
Jump to solution

The vSan standard behavior is to keep the disks as independent.  Its not supportet to use something like HW RAID underneath .

- If you loose a capacity disks and have FTT=1 than you have a copy of your data on another Host. The system will creating a new copy of the data so the level of protection is back

- If you loose a caching/buffer disk than the complete diskgroup goes offline. As long as you have FTT=1 you have a copy on a host which is part of different FT Domain

There is erasure encoding which offering something like RAID5/6 to give better space efficiency. But this depends on vSAN License and also a higher number of nodes is needed.

With vSAN the "protection" is more on Host rather than on a single disk level.

- If you use Dedup and loose a capacity disk than also the entire diskgroup goes offline

Regards,
Joerg

Reply
0 Kudos
b1izzard
Contributor
Contributor
Jump to solution

Thank you both for the explanations.  Very very helpful.  These were the missing pieces for me:

'..if you lose that particular disk due to a failure, the now missing replica is rebuilt from the other copy onto another available disk on the node that lost a disk - losing a single disk doesn't fail the entire Disk-Group unless it is a Cache-tier device or it is a deduplication-enabled cluster'. 

and also 'The vSan standard behavior is to keep the disks as independent.  Its not supported to use something like HW RAID underneath'.

Now it is clear.  Thanks again!

Reply
0 Kudos