2011-03-07 11:35 AM
How are volumes created within an aggregate when that aggregate has 2 RAID groups of different sizes? Currently we have 1 RAID group with 14 data plus 2 parity and another with only 4 data and 2 parity.
I'm concerened that volumes that have part of their data on the smaller RAID group would have poorer performance unless NetApp only writes a fraction of the data on the smaller RAID group.
2011-03-07 11:46 AM
All 16 data disks will help with IOPS in equal manner. WAFL will take care of striping for you. Remember RAID is more for data protection than for IOPS.
Without being able to document it I reckon you re fine.
2011-03-07 12:25 PM
Thanks for the response. I still would like to more in detail how the writes are spread out on the aggregate. Are the RAID groups in the aggregate created so that a volume is striped across them? Is it like JBOD? In my example, I have 14 data disks on one RAID group and 4 on the other for a total of 18 data disks. That would mean, in order to spread evenly, my disk would have to be laid out so that 7/9 (14/18) blocks are on the first RAID group and 2/9 (4/18) are on the second. Does WAFL realize this when volumes are created? Can it *see* into RAID groups for the number of disks contained within?
2011-04-03 02:09 PM
You basically need to find the basic TR (Technical Reports) on the WAFL filesystem (TR-3001, I think) to get a more detailed explanation.
Fundamentally, a flexible volume is simply a storage abstraction. Data from all volumes in an aggregate are simply "mixed" together on the disks. When you add disks to an existing aggregate, WAFL tries to fill the new disks up with "new" data until they are just as full as the existing data disks. This can often cause performance problems as the new disks will be "hot disks" . "Reallocation" needs to be performed on all of the volumes in the aggregate to restore a balanced IO picture.
The rest is in the TR(s).
2014-08-27 11:06 PM
it looks like 6 disk have been added to your aggregate without changing the default raidgroup size of 16 and so you ended up with this misbalanced configuration.
as of the Storage Subsystem Technical FAQ, the raid group sizes should not differ by more than one disk:
"Drive deficiencies should be distributed across RAID groups so that no single RAID group is deficient more than a single drive."
and if I remember correctly, the smaller raidgroup must have at least half the disk count of the bigger raidgroup. but I can't find a doc about that right now.
so I think your configuration is not even supported by Netapp.