Hi guys,
I've been mulling over migration strategies, and babysitting a migration from a F520 with four shelves of 9gb disks to an F740 with three shelves of 36Gb disks.
It's a pain.
ndmpcopy works great for level 0, but has had problems with level 1 copies in my experience. So I'm using rsync 2.4.6 to do the level 1 copies of data to keep things upto date. But it tends to hang, so I have to babysit it all the time. Not good.
Now I know I could use snapmirror to make this migration painless, but that would mean that I'd have to dedicate two 36gb disks on my F740 to the root volume, and then make a second volume with the remaining disks, leaving one spare of course. So I'm losing capacity, but I'm starting to think that this is the way to go, if only to keep my sanity.
And I have another migration to do from an F330 to an F740 with one shelf of disks. In that case, I *really* don't want to dedicate two 35Gb disks to the root vol, and the rest are in a data volume.
So what should I do? ------------------------------------
So what are other people doing when they have to migrate 200-300gb of data from an older netapp to an F740? Are you making a small root volume, and then putting the rest of the disks into a new volume so you can do snap-mirror to cut down the hassles?
Are you just doing what I'm doing now, which is ndmpcopy and rsycn?
Are you just shutting down and doing a dump/restore or level 0 ndmpcopy and to hell with the downtime?
Or is there a secret source of 2gb FC-AL disks that I can put into an FC9 shelf for my root volume?
What I'd like to see from NetApp. -------------------------------------
With the growth in size of disks, I think it's time to re-think how the root volume is managed and configured. The NetApps are there to serve data, and should be even more optimized for that role. But I should not have to sacrifice disk space to make data migrations work smoothly because I need to dedicate two of my large data disks to my root volume, which only takes what, 1gb at most? Heh, I just looked at /etc/ on my F740, it's only 58mb of stuff.
It would be great if NetApp would start putting in a pair of PCMCIA slots on the front panel to hold the root volume. Each slot would hold a solid-state 120mb (or bigger) disk, which would only do the root volume. No moving parts, it's mirrored, etc. Nice and simple. Heck, they could just keep using the WAFL filesystem on there and take the hit. I'd even go for *three* slots so I can keep a hot spare online.
Then we'd have a dedicated data volume using all the big, slow disks for my data storage. Instantly, you'd make migrations of data much faster and easier since we would cut around the snapmirror limitations of using the root volume as the destination.
Another silly idea. --------------------
Or how about the ability to snapmirror qtrees to a root volume on another system? That would be great too! And no need to change the hardware setup as I propose above, it would just *work* right out of the box.
Summary ----------------
Please Give me feedback on my migration, I'm so darn tempted to just lose two disks in my root volume and stuff the rest into a data volume. To hell with the wasted space, my time is more costly.
Thanks, John John Stoffel - Senior Unix Systems Administrator - Lucent Technologies stoffel@lucent.com - http://www.lucent.com - 978-952-7548