Possible solution to resizing an aggregate

[ycdma]

I have a 2 node FAS270C cluster with 14 disks on each node

One node is configured as 14disks with dual parity and no spare. it has only root vol on it.

I need to resize the aggregate down to 2 disks (1data and 1 parity) and use the rest as apare disks.

I wonder how it can achieve without downtime? I need to move the vol0 out and recreate the aggregate and copy the vol0 back?

Any steps/procedures will be appreciated. Thanks

[chriskranz]

Hi David,

I was about to return your email, but I'll simply reply here.

Unfortunately as you may already be aware an aggregate cannot be shrunk. Additionally, moving vol0 requires a filer reboot for the filer to fully recognise it as the root volume, all running system services point to the root aggregate in some way or other, so you need to shut everything down for it to get re-registered in a new location.

So simply, you cannot do what you require live without any down time.

The only way to get your aggregate smaller, is for you to have some extra disks and you create a new aggregate to locate the vol0 onto. A minimum of 2 disks would suffice (RAID-4) if you were just moving it temporarily. You would then have to remove the old aggregate, zero the disks, and then you could re-assign the disks from one controller to the other as needed.

Unfortunately this whole process is relatively disruptive. You can destroy and aggregate and re-assign disks hot, but not move the root volume around, which will be your main challenge. You could minimize this to a single reboot, and so system downtime would be maybe less than 1 minute.

[ycdma]

The cluster node B that I am going to recreate the aggregate have no data on it except the root vol and I have copy the vol0 data on to a backup volume in node A. I am able to reboot node B. Can you briefly tell me the steps that I need to accomplish to recreate the new root vol in node B? Do I need to disable the cluster?

Thanks. David Ma

[chriskranz]

if you have no data on Node A and you need to rebuild the aggregate layout, then you can probably do a 4a on the system. As you already have the root volume copied to Node B it shouldn't be too tricky to copy back the config. Roughly from the top of my head, the steps would be as follows...

1. Disable the cluster
2. Run "license" on the Node A and take a copy of all your license keys (saves a little time later)
3. Connect a console cable to Node A
4. Reboot Node A and when prompted, hit CTRL+C for special boot options (not to break the boot sequence).
5. From the maintenance mode menu, choose 4a
6. This will remove any aggregate configuration and data completely off the disks on Node A. This will take awhile as you are reformatting the disks.
7. Once the 4a has been completed, the system will reboot and you'll run through a setup wizard to configure the filer. Give it a basic configuration so you can get on the network and so on.
8. Re-add all your licenses.
9. Copy the vol0 backup from Node B back to Node A. You'll have to copy it to a different location (maybe vol0_new).
10. Change this new volume to be the root volume "vol options vol0_new root".
11. Reboot Node A once again. This will boot into your original config
12. Delete the second vol0 that is no longer being used, and rename vol0_new if you want.
13. Re-enable the cluster.

Remember this is a pretty destructive method and you'll lose all data on Node A, so make sure you are happy with this first. You'll end up with a 3 disk aggregate (RAID-DP) on Node A and your original root volume back in place, additionally 11 spare disks for your disposal!

Node B would not need to be rebooted at all, so if you are serving any live data from there, you can continue to do so.

Let me know if you need any further info or clarification. Good luck!

[ycdma]

Thanks for your info. Couple more questions:

1) when I copy all the data, will they copy the origingal 14 disks raid group configuration back? How it comes with a 3 disk aggregate with DP?

2) Will the reboot options same for ONTAP 7.0.2?

Thanks,

DMa

[chriskranz]

The only thing you should be taking a copy of is vol0. A volume exists within an aggregate, regardless of the aggregate configuration. So copying the vol0 back onto a 3 disk aggregate that was previously on a 14 disk aggregate will cause you no issues with vol0 at all.

All this should be true of all ONTAP 7.x releases.

[chriskranz]

The only thing you should be taking a copy of is vol0. A volume exists within an aggregate, regardless of the aggregate configuration. So copying the vol0 back onto a 3 disk aggregate that was previously on a 14 disk aggregate will cause you no issues with vol0 at all.

All this should be true of all ONTAP 7.x releases.

[ycdma]

so the default setup is a 3 disk aggregate after the reboot?

And I can use ndmpcopy to copy the root vol form node B to a new vol in node A?

[chriskranz]

After you've performed a 4a from maintenance mode, yes, you will have a fresh 3 disk aggregate. And yup, then you can use NDMPcopy to copy the vol0 back across.