Here's the idea we like the most round here, but it takes investment.
Be persuasive with your salesguy and arrange for a newer box, F520 say, spec'd
to cope and arrange for a decent overlap between setting up the new system and
taking the old one away - if you get some kinda upgrade deal that is. Sounds
like you could probably use a nice stable F330 elsewhere anyway so you might
keep the old box too. 8)
Use NDMPCopy 1.1 (with DOT4.3 I think) to do a big copy of all the data when
things aren't too exciting on the network or the filer. Either that or use a
full backup on tape to copy the bulk of the data. (presupposing you take local
backups like we do, which most people appear not to for some reason)
When you get your chance to actually do the switcheroo take the F330 offline
wrt the clients, do a final incremental to the new box over the network, make
the new box look like the old one to the world (and make the old one look like
something else...too) and do a bit of testing from your clients.
If all seems stable at that stage then run with it in full-production style. If
problems crop up, you can breath a sigh of relief that all the old data is
still safe and instantly accessible on the old filer and deal with things as
they happen.
If you've gotten through your first, say, couple of weeks then you can have
high confidence that the new filer won't give out on you.
Certainly this is pretty much how things have worked when we've upgraded our
filers here (and believe me, I've now upgraded 8 in different and *interesting*
ways). The real enabler for us to do the job without threatening the data (the
important stuff after all) was that our NetApp engineers let us sit on kit when
things looked dicy and gave us excellent support in putting the boxes together
and dealing with the bedding-in problems. I could tell you tales of the late
nights they've put in completely stripping down and rebuilding cabinets full of
kit...
All you need is the option to backout and suddenly going forward doesn't seem
quite so nerve-wracking. Plus the visible downtime of the data can be cut to
minutes if you have the details down pat.
Hope this helps.
On Nov 15, 9:39, Daniel Quinlan wrote:
> Subject: Upgrading from an F330
> One of our filers, an F330, is very overloaded (80-100% CPU usage and
> 1000-1500 NFS ops/sec for most of the 24 hour day) and is always very
> close to being full. Both shelf pairs are full of 4 GB disks. We
> also have an F540 and an F630, which are far less loaded.
>
> Interestingly, the F330 is the most stable filer we have had: no disk
> failures in over a year, no major problems. I wish our other filers
> were so rock solid. *sigh* Our other two filers have had so many
> problems, I can't remember them all: failed disks, failed ethernet
> transceiver, bad shelves, SCSI problems, etc.
>
> Unfortunately, the data on the F330 is some of most critical data at
> our company. More than a few hours of downtime is *not* an option.
>
> I read Tom Yates's "migrating an F330 to a mixed-disc configuration"
> messages, so I'm not really inclined to try that. Are there any other
> options for fixing the performance and disk space problem that you
> would recommend?
>
> - Dan
>-- End of excerpt from Daniel Quinlan
--
-Mark, TSG Unix admin and support, int 782 4412, ext +44 1438 76 4412.
... an Englishman in London ...