On 10/08/97 09:10:10 you wrote:
We have "raid swap"ed almost three shelves worth of drives into an F330 at one time. The AE recommended this method as being faster than installing all the drives at config time and building that large file system from scratch. This worked fine, but still takes a significant amount of time.
I think by "at one time" you still mean one at a time (a raid swap for each drive).
We later added a fourth shelf and "raid swap"ed all those drives in at once. The filer was only off the net for the length of time it took to halt/connect the SCSI cable/ reboot. A few hours later, after the raid swap completes, all the additional disk space shows up - it does not show up one drive at a time.
I don't quite understand this... if you added them while the filer was down, you didn't "raid swap" anything. And they show up immediately, not a few hours later. I think perhaps you are talking about *raid add*, i.e. actually adding the drives to the filesystem. In that case you can add them individually one drive at a time (and the filer would get the space on drive at a time), or you can raid add them all at once, in which case they'll all finish at about the same time.
Some comments regarding DLT drives - I have heard that you want to keep them "streaming" (feed with data) or the drive has to stop, wait for data, backup, reseek to the stop point, and start again. This applies to both 4000/7000 series. You pay a significant speed penalty for dropping the drive out of streaming mode. Our word from Netapp is that for an F230, they sell the FW/DIff card and it works, but it is not officially QA'd in the system, but soon should be.
Yes, streaming is important, but my point was I don't think a 7000 non-streaming is going to be much slower than a 4000 streaming. But, I could be wrong on this. In any case, there are *plenty* are start-stop behavior in a Netapp dump even locally, so I don't think you can avoid it.
Bruce