Need advice/options on using leftover devices

[jamesNJ]

Netapp community hello!

I need some advice on what to do with and odd disk config I was left with.

I have a FAS8300 that has has 3x DS212-12 shelves each fully populated with 8.89 tb drives. As an aside, it seem like these 8.89tb drives are classified as “10tb or higher” capacity devices.

The implementation left me with 2 RAID-TEC aggregates (aggr1, aggr2) that each use 1 shelf of 12 drives. These aggregates also use partitioned drives in order to host the root aggregates. That leaves me with 1 shelf of 12 drives completely unallocated.

I’d like to create a new local tier to use these devices. Ideally I like 2 similar size aggregates using 6 devices each. However, it seems like for the class of device I have, I can only use RAID-TEC which requires a minimum of 7 devices.

So I need to know what my real options are. Ideally, I’d like to eject 1 device from each of the existing aggregates so that I have my 7 devices. Or, find some way to get around this 7-device minimum, perhaps use RAID-4 instead. I’m not sure if either approach is feasible.

If not, I could just create 1 aggregate with all 12 devices, but I’d have to put that onto one of my heads and there would be greater chance for unbalanced I/O load. I’m not sure how much I should really worry about that as none of these aggregates are expected to be heavily loaded.

Any suggestions appreciated, thanks

[unixnation · Accepted solution]

Hi there,

If you're worrying about unbalanced I/O to the controllers, I wouldn't. If there is an I/O concern in this configuration, I think it would be more the small number of comparitivily slow devices in the aggregates - which I assume is actully fine otherwise you'd already have a problem?

I don't think you are going to be pushing the controllers themselves based on what I see on my FAS8300. If it was me, I'd be tempted to increase the size of the existing two aggregates to give more capacity and spindles to serve requests. It seems a waste to create another aggregate + incur 3 more parity disks when you have so few to start with.

Cheers,
Steve

[jamesNJ]

Thanks for the reply; I don't disagree. I have encountered bad I/O issues on small aggregates in the past.

The reasoning for multiple aggrs is to maintain flexibility. The array is shared between teams for dev/qa work and really shouldn't be getting hit hard with I/O (except when someone decides to try performance/load/scale testing). Once you put all the drives into an aggr, pulling them out for some other project is unlikely, or at best a very painful exercise.

I think in this case I'll just put the remaining disks into 1 aggr and live with it until I have an actual concern to deal with.

Thanks

[unixnation]

Sounds like a reasonable compromise to me!

I totally agree on it being a painful excerise if you have to remove/repurpose drives that are part of an aggr. Still waiting for the "remove storage pool from hybrid aggregate" feature here...

Cheers,

Steve