ONTAP Discussions

Highlighted

When adding new disks to an aggregate can all disks be used for data?

We have aggr1_Node03 that contains 22 disks (with 2 hot spares).  Those disks are split into 2 RAID groups (each with 11 disks) and there are 2 parity disks per raid group.  If we added another DS212C with the same size drives that are currently in the aggregate, can we just increase each RAID group size to 17 disks and add 6 more data disks to each group or will we be required to also add more parity and spare disks?

8 REPLIES 8
Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

No, the parity 'disks' will remain '2' per RAID-DP raid-group (No more parity tax). You can easily expand it 17 (with each 6 additional disks). In general recommended range of RAID group size is between 12 and 20,  and adding '6' together sounds  good.

On the side note: With larger RAID groups (disks), only caveat is that - it will take longer to calculate parity because there are more blocks to calculate parity for.

View solution in original post

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

What is the best practice regarding RAID group sizes in the same aggregate?  As mentioned, we currently have two RAID groups that are both 11 disks.  One option would be to get a DS212 (with the same size disks) and add 6 disks to each group.  Another option we were looking at was to get two DS212 shelves with the same size disks, but then we would have two new RAID groups of 22 disks in the same aggregate with the RAID groups that are 11 disks.  Does that present any issues?

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

I have seen configurations of mixed RG sizes. I don't believe we have a specific best practice, but you didn't mention if SSD or HDD. That makes a difference too.

 

Edit: After pushing send, I found this: https://docs.netapp.com/ontap-9/index.jsp?topic=%2Fcom.netapp.doc.onc-sm-help-900%2FGUID-345CE3A2-B68D-450B-86BE-D144698FEF76.html&lang=en

The recommended RAID group size is 12 disks through 20 disks for HDDs, and 20 disks through 28 disks for SSDs.

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

We were looking at adding 8 TB SAS drives just like the ones that we currently have.  With the setup that we decided on we are going to have RGs that are the same size.  Thanks for the help!

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

We now have added 24 8TB disks in one DS4246 shelf to the cluster.  Is there a best practice in terms of whether it is better to add disks to the existing aggregate (same size 8TB disks but a different model and on a different loop) or to just create 2 new aggregates?

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

It really depends on your needs, so we don't publish a specific best practice that I am personally aware of. Usually the larger the aggregate (in terms of # of data disks), the better the performance from your spindles. You wouldn't want to create an aggregate if you only had 3 data drives when you could add to an existing aggregate and have 20 drives.

 

If you grow an existing aggregate, make sure you run reallocates.

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

If we added them to the existing aggregates, we would end up with 3 RAID groups of 11 disks (33 disks total) in each aggregate.  

 

We just weren't sure if they were performance concerns mixing X318_HARHE08TA07 and X318_HLBRE08TA07 disks that are in different shelves on different loops.

Highlighted

Re: When adding new disks to an aggregate can all disks be used for data?

No. Just make sure they have latest firmware. They both are Hitachi drives but NA01 on the Hitachi Libre drives really improves performance 3x over NA00. I was involved on that escalation/bug that resulted in the release of NA01.

Check out the KB!
Knowledge Base
All Community Forums