Replace one FAS3020 controller to FAS3240

[umonteiro]

Hi everybody,

Which correct procedures to replace one FAS3020 controller to FAS3240 with minor impact possible ?

The FAS3020 is HA (cluster mode) in takeover state but i have to replace for a single controller FAS3240.

Any suggestion or documentation will help me.

Tks All!

[andrc · Accepted solution]

Downtime will be necessary because you can't have different controller types in a cluster. Carry out the following:

Issue `cf giveback` to bring up the node currently in takeover.

Confirm normal status of both nodes.

Check if software of hardware based disk ownership is in use.

Ensure the controllers have different root volume names e.g. /vol/rootvol_filera & /vol/rootvol_filerb.

Note which controller's config you want to keep on the FAS3240 controller, edit the /etc/rc file on that controller to reflect any hardware changes on the new controller e.g NIC numbering.

Make a copy of all config files from /vol/rootvol/etc

Note all LUN serial numbers if applicable.

Issue `cf disable` and power down both controllers.

Remove the FAS3020 controllers, install the FAS3240 controller and cable the shelves.

Make a console or RLM connection, power up and boot to Special Boot Options Menu (Hit ctrl/c when prompted) choose option 5, Maintenance Mode Boot.

Issue `disk show all` and check disk ownership, if required issue `disk reassign` to assign disk ownership to the new controller.

Offline the controller root volume you don't require and ensure the root volume you do require is online. Confirm all other aggregates/volumes are online.

Issue `halt` to get to the loader prompt, then issue `bye` to reboot the filer.

If everything's been done correctly the filer should boot using the correct root volume and /etc/rc file and you should be able to login.

[umonteiro]

Excellent procedures but the FAS3020 one node in takeover because the other node is DOWN (hardware fault).

Basically we intend change this one node FAS3020 for other FAS3240.

Change anything in your recommendations in this case ?

ps. See the image in attach

tks a lot!

[andrc]

In that case it should be pretty much the same although without the cluster giveback steps.

If software based disk ownership is in place then you'll still need to reasssign the disks in Maintenance Mode Boot, but be aware that both root volumes will be recognised and the filer won't know which one to use so one will be marked foreign and offlined. You might also run into an issue that if both controllers have aggregates/volumes with the same names then one will be marked as foreign and offlined.