2012-07-08 06:16 AM
I have to configure a fas2240, that has cf und syncmirror_local licensed on it, and 24 X 300GB internal disks. I'm very insecure about the storage configuration for it, in order to have the best of ha and at the same time not loosing to much of disk capacity. What I've found out by netapp setups recommendations is "Disks or array LUNs in the same plex must be from the same pool, with those in the opposite plex from the opposite pool." Nothing else, but I am assuming that the storage amount for both controller must be at least of the same size (for the mirroring) and I would configure this filer like that:
contrl 1 - 1 spare, 1 aggr (for root and everything else) with raid4, 11disks
contrl 2 - 1 spare, 1aggr ((for root and everything else) with raid4, 11disks
what means that I would have just 10 disks for real data, and if one calculates each disk with about 70% of real data capacity, I am ending with just about 2TB for data. Is that correct? Could I do it somehow better? Has somebody any other idea about configuring it in order to have more data capacity available?
Or should I just forget syncmirror, and then I could configure one controller (the one for the real data) having more disks than the other, raid-dp (and the other with raid-4)?
Thanks for any correction of the above thoughts or other ideas for the storage design.
2012-07-08 07:52 AM
Syncmirror is unrelated to HA; you do not need this license in your configuration and should remove it. Using syncmirror will reduce available storage by half. It is used to additionally mirror between two aggregates on the same filer head.
Most people (me including) would recommend using RAID_DP. If space is at premium, using RAID_DP without spares can be considered; at the end it depends on how fast failed disk would get replaced.
Each head must have own disks which means you cannot do much better than you described. Distribution of disks between heads is determined by storage needs, but each head must have aggregate which must have parity disks and should better have spare as well.
2012-07-08 02:43 PM
thanks, I'm getting there.
I thought the data from one head (controller) would be mirrored to the other one . And because of that I was thinking of raid4 and not dp (never had a failure of 2 disks at same time). Your alternative, withou syncmirror, would be then for me one controller with dp (maybe the one with the important data) and the other one raid4, holding just the root volume. The ha hier would be just for a controller failure.
Storage something like that:
contrl 1 - 2 spare disks+ 1 aggr (for root and the important data) with raid dp, 19 disks (2 for parity, 15 for data) ~4,7TB
contrl 2- 1 spare disk+ 1 aggr just with root, raid4- 2 disks (1 parity and 1 for data)- total of 3 disks
And of course the disks are not 300GB, but 450GB
I've forgotten to mention that this filer is going to be used mainly as iSCSI LUNs for wmware, and in view of that, whether it makes sense to use syncmirror at all.
Sorry, it seems that it does not make any sense to use syncmirror with just internal disks (because the aggr will be mirrored inside of a controller)...
I will do without it, like above.