Provisioning Question for New 'Very Large' Volume

[bsnyder27]

We have a request to add a very large amount of data to our environment. I'm looking for recommendations on whether or not to put it into one volume or break it up.

Hardware: FAS8200 HA pair w/ (2) aggrs in DS224C and (2) aggrs in DS2246 shelves. All 900GB SAS drives

Request: Initially, 30TB of data to be moved. Not certain upper growth in a year's time but let's go with 50TB.

We have 4 aggrs that are practically all the same speed disk. Here is the free space on them

1. 32TB
2. 33TB
3. 40TB
4. 23TB

Given 50TB is not even possible on any single aggregate today, that option is out leaving us with splitting it into n volumes. The unknown here for me is an Infinite Volume as we've never provisioned one.

If we were to go this route, are there any gotchas given the following criteria?

• coexist with normal flex vols on the aggrs it spans with the understanding it needs to be in its own SVM
• has a snapshot schedule
• is snapmirrored to another FAS

Just looking for the best way to approach this given the storage we have available today. Thanks for any suggestions.

[colsen · Accepted solution]

If you're running ONTAP 9.1+ and will be serving this up as NAS, I'd take a serious look at FlexGroups:

http://docs.netapp.com/ontap-9/index.jsp?topic=%2Fcom.netapp.doc.pow-fg-mgmt%2FGUID-A304BBC1-C00C-4E7A-989E-7C5A0E505146.html

Definitely some cool technology.  You could also non-disruptively move some volumes around to free up space in a single aggregate - more work but tried-and-true.

Hope that helps,

Chris

[AlexDawson]

Flexgroups are definitely a good option here - but wondering why you didn't go for all the DS2246's on one node, and all the DS224C's on the other, and make the aggregates as large as possible? This would be what I would suggest to most people, as it avoids the problem you're seeing now, and we support max IOPS QoS on volumes/SVMs, in order to partition performance at a higher and more flexible level.

[bsnyder27]

So unless it's still a bear to move a root volume in CDOT, and assuming all of our current data can fit on the (5) DS2246 shelves...it appears I might get a 4-5 extra TB from doing just what you suggested. There are (7) DS224C shelves. All 900GB SAS drives and all shelves fully populated.

What would be the recommendation for spare disks with this configuration? Just 2 spares for each type of disk suffice?

Would this be a good idea after the fact? Running OnTap 9.1

[AlexDawson]

With 9.1, moving the root volume is a lot simpler than pre-9, so it shouldn't be too big of a problem.

Our minimum spares recommendation is 1+1 per 100 drives - so with 7 x DS224C I'd go for three drives spare, and 5 x DS2246s, I'd go for three as well.

This is still an option with a mix of shelves on each system, ie, only three spare per controller, but it would be messier. It is not recommended to mix the 12G and 6G drives in the same aggregate. 

So in summary, yes, I would split the two disk types to one per controller, and have the largest aggregates you can and use QOS to isloate workloads.