2014-07-11 02:34 PM
One node is using just SAS disks, other node is strictly SATA disks. DS4243 disk shelves.
The node running the SAS disks (4) DS4243 shelves has been running the disks at 80-90% busy for the past 3 months and we've been dealing with performance issues. We just purchased a 5th DS4243 shelf and added it to the aggregate today. RAID group size it as 23.
The way I understand how NetApp works all writes now will automatically be flushed to the new disks until they fill up equally to the level of the entire aggregate then writes will stripe across all disks. -> Is this correct?
But my main question is what would be the best thing to do at this point to maximize the new disks and help to reduce the busyness of the other disks in the aggregate?
2014-07-11 06:55 PM
Well, i'm always a fan of even number raid groups. Even though netapp says you can go to 28 disks in a sas rg, that risk profile is too high for me.. 23 is pushing it a little as well.. Highest i've gone was 24, in isolated cases.
I suggest you read the following document - reallocate best practices tr-3929
I still don't understand why netapp never thought to re-allocate the entire aggregate when adding disk. That would be a big win. It's a tedious process, but if you want to stripe the data across all your disks, this is what you will have to do
2014-07-12 08:16 AM
What's your reasoning for even RAID groups instead of odd? And why are you worried about the size, with dual parity and hot spare there should never be an issue.
Have you done reallocation? If so what type of performance hit have you seen and time frame for the job to complete?
Thank you for the input.