question about growfs, aac raid, enlarging by adding disks
Scott Long
scottl at freebsd.org
Tue Apr 27 19:05:44 PDT 2004
Don Bowman wrote:
> this is perhaps a naive question, but i
> couldn't find the answer.
>
> if i have e.g. a 3-disk raid5 using aac,
> and i add another disk, can i increase the
> size of the logical disk using the aac bios
> (or ideally aacli), and then use growfs to
> increase the size of the filesystem? Can
> this be done if i obey some restrictions,
> e.g. resize the last filesystem on the partition?
>
> I would like to boot from the array, have swap
> on it, and have a large partition that uses
> the remainder of the space. I'm assuming i would
> lay it out so that it goes root first, then
> swap, then the large filesystem.
>
> The adaptec raid bios seems to indicate i can
> grow by adding a drive, but only under windows.
> The manual for it has no mention of this. The
> datasheet also indicates i can increase size
> 'on the fly'.
>
aaccli will let you increase the size of the container.
If you are using Windows, then there is magic that will
resize the partition table and the NTFS and/or FAT
filesystems. I can't remember if the magic is in the
driver, aaccli, or firmware, but it definitely only
applies to FAT and NTFS _filesystems_. Growing the
container is done in the firmware and is independent
of the filesystems, MBR/slice table, etc.
> Does vinum enter into this? ccd? Since the raid
> is done in hardware i wouldn't think so.
>
neither vinum nor ccd apply here. A true LVM might
be nice, but vinum doesn't provide the kind of LVM
services that would help here.
> It would appear that i might be able to add
> a disk, convince the raid to increase in size
> somehow, rerun fdisk to increase the size of
> my partition, rerun bsdlabel to increase the
> size of my slice, and then run growfs.
>
I agree with this. Of course, nothing beats a good
backup first =-) Let me know if you need help with
the aaccli magic.
> this is all on 5.2.
>
> Anybody else ever done this?
>
Years ago I tried it and was somewhat successfull, but
it was more of a proof of concept than anything else.
> --don
Scott
More information about the freebsd-current
mailing list