Sorry for going a little off topic...
Have you tried an ICP Vortex GDT6513RS card? I have a server with 9 180 GB
scsi drives (baracudas from seagate) that I'd like to have in a single
RAID 5 volume. I've had some problems getting just a ext2 filesystem on
it, seems like I can only get it to work with a maximum of 3 drives in a
single volume. Would love to make it run with an XFS filesystem...
Can I ask what type of kernel changes you had to do with the 3ware to make
it work?
Shawn
On Mon, 18 Jun 2001, Justin Tripp wrote:
>
> I have used both the 3ware and Promise Raid solutions. Their are two
> Promise RAID cards. Only one will work with linux and it is a software
> raid solution. Promise only supports RedHat 6.1,6.2 and Suse 6.2, and
> only the basic kernels in those releases. So, if you are not interested
> in XFS, then you could try them out. It is likely, though, that you could
> get the same performance (and cheaper), but just buying a additional
> promise IDE controller and running the kernel's software raid on top of
> that. And then you could run any kernel you desired to -- plus XFS. IDE
> software RAID is not likely to have too great of performance. (FYI
> Promise's kernel driver is not open source, only part of it is and I never
> did get them to return any emails I sent them about using the card under
> linux).
>
> 3ware, on the other hand, is a hardware raid. As far as your computer is
> concerned you have a SCSI controller hooked to a pci slot. The overhead
> of IDE disks is not seen by the processor since it only interacts with the
> Raid controller and cannot see the individual disks. I have a 3ware
> controller with 4 40G disks running RAID 5 with XFS as the underlying
> filesystem. It works quite well.
>
> 3ware did have problems with the firmware and RAID5, but these have been
> fixed and can easily be obtained from their website. The problems only
> occured if the RAID ran in degraded mode, and if one did not have a disk
> failure, then the problem never occurred. (Phew). The kernel driver
> works well in the 2.2 kernels, but 3ware is not too keen on helping you
> run in 2.4. Now that some distributions ship with 2.4, they may be
> changing their minds, but... Nonetheless, their driver is open source and
> is part of the kernel tree. The driver I am running is cut out of the
> 2.4.5-ac5 patch and applied to a 2.4.5 cvs version of XFS. I have not had
> any problems despite the fact that XFS seems to exercise the underlying
> hardware more than ext2 does.
>
> I believe the 3ware card is a reasonably cheap way to get a 100+G raid.
> But I did have to do some light kernel hacking to get everything working
> reasonably well.
>
> .justin.
>
> On Mon, 18 Jun 2001, Seth Mos wrote:
>
> > At 09:13 18-6-2001 -0400, Andrew Klaassen wrote:
> > >On Mon, Jun 18, 2001 at 03:55:36PM +1000,
> > >Robin Humble wrote:
> > >
> > > > Around the time of the 2.4.3 kernel we used XFS over software
> > > > RAID5 for a month or so before a disk died and we didn't
> > > > bother replacing it - we've been using 420G (7 disks) of RAID0
> > > > since with zero problems. RAID5 seemed ok and we sorted out
> > > > any initial performance problems as we found them with the
> > > > super-responsive XFS people on this list.
> > >
> > >7 disks? I'm curious: SCSI or IDE? (We're looking into a
> > >Promise or 3ware card to allow us to put lots of IDE drives in a
> > >box and run software RAID over top, and were wondering if anyone
> > >else has had experience with these cards+software RAID+XFS.)
> >
> > IIRC the 3ware driver is a bit shakey. The 3ware card is a hardware raid
> > solution and not a software one. If you use promise cards you will probably
> > use software raid.
> >
> > The issue with the 3ware raid was firmware related. If the controller was
> > missing a disk and running in degraded mode file system corruption could
> > and would occur.
> > Other then that it is a fine IDE raid controller.
> >
> >
> > >Andrew Klaassen
> >
> > Good luck
> > --
> > Seth
> > Every program has two purposes one for which
> > it was written and another for which it wasn't
> > I use the last kind.
> >
> >
>
>
|