Double aggr size by upgrading disks, without destroying/rebuilding aggr?

[WSANDERSATFLEXERA]

A Netapp support engineer just told me I could double the size of an aggr by replacing all its 1/2 TB disks with 1 TB disks, without destroying and rebuilding the aggr.

This might be feasible for us since we have an aggr that started out years ago as all-1/2 TB disks, and over the years Netapp has shipped 1/2 TB disks to us to replace 1/2 TB disks that have died, and/or people have bought 1TB disks and mixed them into the aggregate. The aggr is now about 2/3 1 TB disks, of course the 1TB disks are all downsized to 1/2 TB.

The Netapp engineer says that if we: First step - upgrade all the parity disks to 1 TB, and Second - upgrade all the rest of the disks in the aggr to 1TB, the aggr will start using the full size of the disks and essentially the size of the aggr will (magically!) double.

It might not work in our case - the aggr is already 46 TB with the disks at 1/2 capacity, and there is a 75 TB limit on aggr size on our filer (FAS3140/Ontap 8.1.2).

Considering the cost, it only makes sense in a few cases, like when aggrs are fairly small and most of the disks have been upgraded by disk failures over the years.

The other caveat is that we were told it works only for the entire aggr, not just one raid group at a time.

Has anyone done this? Does it work?

Thanks,

w

[aborzenkov]

That's very interesting statement, because it was not possible in the past 10 years ...

You can add larger disks to aggregate as new RAID group; but you cannot replace disks with larger disks and use extra space - disks will be downsized to match original small size.

May be NetApp engineer meant that you can create aggregate using new disks and move volumes to it? That's obviously possible.

[resqme914]

I've always wanted to do this but it seemed impossible.  It would be great if it could actually be done but I tend to agree with Aborzenkov and I'm doubtful.

I guess this could easily be tested.  I have a filer with a whole bunch of currently-spare disks of varying sizes.  Just have to find the time.

[WSANDERSATFLEXERA]

Thanks, all. FWIW, we do have some raid groups where the 1/2 TB "dparity" disk failed, as well as some others, and the dparity disk was then *not* downsized. The "dparity" disk was not downsized, and the "parity" disk was:

RAID group /aggr1/plex0/rg2 (normal, block checksums)

      RAID Disk Device          HA  SHELF BAY CHAN Pool Type  RPM  Used (MB/blks)    Phys (MB/blks)
      --------- ------          ------------- ---- ---- ---- ----- --------------    --------------
      dparity   1a.11.20        1a    11  20  SA:A   -  BSAS  7200 847555/1735794176 847884/1736466816
      parity    1a.11.18        1a    11  18  SA:A   -  BSAS  7200 423111/866531584  847884/1736466816
      data      1a.11.0         1a    11  0   SA:A   -  BSAS  7200 423111/866531584  847884/1736466816
      data      1c.02.19        1c    2   19  SA:B   -  BSAS  7200 423111/866531584  423946/868242816
[etc - mix of 1/2 and 1 TB disks]