RG Size 2TB Disks

[nsitps1976]

With a FAS 2020HA with 12 2TB drives is this a valid config:

1 Spare

1 RG =  11 disks, 2 parity and 9 data

All assigned to one node, the other acting purely for failover.

I do not want to lose 6 disks by having a RG per node.

Thanks

[aborzenkov · Accepted solution]

To allow failover between heads in active/active pair each head must be up and running; and to be able to run, each head needs at least root volume. Your configuration leaves second head as cold spare. It could be acceptable if you can afford downtime for replacement (I had customers with single FAS250 or R200), but it is pretty expensive spares stock IMHO.

[nsitps1976]

Oh!

So what is the best way to configure this so as to lose as little disk as possible?

Two RG of 5 disks using Raid4 (4 data and 1 parity) - One hot spare per node ?

Also, in the cold senario you described what would be the process to swap the cold controller for a failed active controller?

[aborzenkov]

Various replacement procedures are described on NOW site.

I really can’t say which is the best. In similar situation several years ago customer opted for hot standby with 2 disks in RAID4 for root and no spares on one controller, but they were 72GB disks; I guess you would not like to lose 4TB worth of space. If you do not absolutely need large contiguous chunk your suggestion looks reasonable; although I’d rather go for RAID_DP without spares.

Search communities, this question pops up pretty often and there were several long threads about it already.

[brucem]

As noted already, you need a root volume for each controller. So best bang for your buck would be a 3 drive Raid-DP on the 2nd controller ( 1+2 no HS)  and the other controller would have a 7+2 (raid DP)  No HS, or a 6+2 DP and 1 HS, your call

[nsitps1976]

I thought it was a requirment to have at least one HS per controller?

How is this configured? Will the system not complain?

Also, can you point me in the direction of the threads re this, I can't seem to find much on my senario.

Regards

[aborzenkov]

Yes, system will complaint but run. In some cases it is possible to turn warnings off - see raid.min_spare_count option.

Similar posts:

http://communities.netapp.com/message/45564#45564

http://communities.netapp.com/message/32426#32426

http://communities.netapp.com/message/42579#42579

[nsitps1976]

So there is no special config required in order to not assign a HS - You simply just use all disks for the aggr ??

[aborzenkov]

Mostly yes, but do not forget about raid.timeout option. I learned about it hard way when my very first NetApp suddenly started to switch off ☺

[nsitps1976]

Please could you explain what raid.timeout option is and what I need to do?

Is it only needed when you do not have HS disks?

[aborzenkov]

NetApp will shut down if degraded RG is not repaired within raid.timeout timeframe. It is to protect you from further disk failures. It is described in Data ONTAP manuals.

[nsitps1976]

Thanks for all the help