2010-05-20 04:15 AM
Hi and welcome to the Communities!
Have a look at these pages:
- If your filer is single-controller - http://now.netapp.com/NOW/knowledge/docs/hardware/NetApp/cs_migration/FAS3xxx/FAS3020-FAS3050_to_FAS31xx_single-controller/GUID-FC68C804-80BA-4F10-84D...
2010-05-21 08:05 PM
Just to echo Radek, I always use those guides even though I do upgrades semi-frequently.
The main page for NetApp hardware upgrade guides is...
2010-05-21 11:16 PM
thank you for link.
Since it is mass production enviromenet .I want to reduce the down time.So i am in process of prepraing doc .I need to find all the steps and risks.I welcome you inputs
2010-05-24 07:26 AM
If it's a cluster (I would assume so), make sure you destroy the mailbox disks after the upgrade. When you go from 32bit to 64bit systems this is required, but I haven't seen it too clearly documented. The cluster will recreate these, so it's not as harmful as the command may suggest (so long as you cleanly disable the cluster before the work and don't have a cluster failover!). Only other point from experience is to make sure you re-install Data ONTAP afterwards. Sometimes you can have such relief after having a nice smooth upgrade you completely forget to do this as you don't notice any immediate issues (not that I've ever done that! ).
Also make sure both systems are on the same version of ONTAP before starting the upgrade. You can downgrade a filer that has no disks by using netboot, but it'd be best to upgrade the existing filer instead. You may want to schedule 2 outages to minimise the overall impact, one for an ONTAP upgrade, one for a headswap.
Otherwise head upgrades are pretty nice and simple. If you are absolutely flawless with the work and you aren't doing any other work at the same time, I reckon you can do it inside of 2 hours. For safety, I'd always try to schedule in a 4-5 hour outage window however. Remember a power cycle is the point where a system is under most strain, so anything can go. If you have 4 hour parts replace, then if you reboot the filer at the beginning of the work, you can be sure that if anything goes "pop" you can get a replacement within the outage window still (definitely something from experience!!!).
Make sure you also map out the network connections. Remember a 3020 will have different port configurations for all your networking, so this will need to be reconfigured when the 3170 comes up. If you have the RC file pre-written before the work, you can make your life easy and just drop the new one in place.
Also connect a console cable to the system, and power the new head up first, make sure it passes all checks. At the system prompt run "boot_diags" and do a full system check. Do this on both controllers in the system. You don't want to be the unlucky minority and have a DOA system (I've had 1 in 5 years, about 300 systems, but it is still a possibility, however minor).