Hi guys,
I'd like to post a summary of what I ended up doing for our data migration this past weekend. With the help of a couple of pointers on this list, it turned out to be quite simple, once I was shown how you can change different volumes into root volumes with just a reboot.
On the downside, NetApp's document "Migrating Data Between NetApp Filers" is sadly out of date, and doesn't try to use some of the tricks that are possible with snapmirror if you have a large enough filer to use as the destination source.
Basically, here are the steps I took, in a quick high level example. The devil is in the details, so let me know if there is anything I left out here.
0. Read and understand the instructions in the "SnapMirror Operational Guide" from the NetApp web site.
- make sure you're at 5.3.x or higher on both src and dest.
1. Build your new destination filer with the minimum root volume, by default it's called vol0.
2. Create a second volume on the destination filer with enough room to hold the source filer data volume. Which in my case was a root volume.
3. Start the initial snap mirror process going, then get a couple of cups of coffee. It took about 6-8 hours to do the initial mirror of a 240+gb volume from an F520 to an F740.
4. Keep migrating data in secondary snapshots. Repeat until your migration window opens.
5. Re-export all filesystems on the source filer volume Read Only to the clients. Do the final snapmirror, I did two because I was concerned about getting it right.
6. Once this final snapmirror has completed, shutdown the source filer.
7. Edit the files in etc/ on the new filer to make sure you have the correct network interfaces, etc defined. This makes the changeover much easier.
8. On the destination filer, bring the destination volume online, then use the 'vol options' command to make it the root volume. The reboot.
NOTE! In my case, both the F520 and the F740 are Alpha architecture, so I did not have to install the proper DataOnTap architecture files onto the new volume before doing the reboot.
This step would expand a bit if you were going from an F330 to an F740, like I will be doing later on at some point.
9. Make sure you're clients are all set. We just rebooted all the Unix clients since it was simpler, and it simplified the checks we had to run later.
10. After a week or two, delete the old root volume on the destination filer and incorporate those disks into the new root volume. In my case, I purposely left 5 other spares around so I would be able to create a more efficient RAID set when I did add those disks back in.
I'd be more than happy to expand this document at some point when I'm not dead tired.
Thanks to all those who pointed out the trick with changing root volumes, what a time saver.
John John Stoffel - Senior Unix Systems Administrator - Lucent Technologies stoffel@lucent.com - http://www.lucent.com - 978-952-7548