Re: Building new Athlon AMD64 Socket 939 or 940 machine

From: Gavin Atkinson <gavin.atkinson_at_ury.york.ac.uk>
Date: Tue, 29 Nov 2005 13:37:14 +0000
On Mon, 2005-11-28 at 13:05 -0500, Martin Cracauer wrote:
> > Hmm, well that's promising.  This is with:
> > atapci1_at_pci0:7:0:       class=0x010485 card=0x5348108e chip=0x005410de
> > rev=0xf3 hdr=0x00
> >     vendor   = 'NVIDIA Corporation'
> >     device   = 'nForce4 Serial ATA Controller'
> >     class    = mass storage
> >     subclass = RAID
> > 
> > I don't know if it makes any difference, but the board has on-board
> > NVIDIA RAID, which I'm considerably less than impressed with.  I'll
> > build one of them without the on-board RAID support and see if
> > hot-insert works.  I don't have the logs (long story, but basically the
> > RAID bios hosed my data) but I think it failed to notice the inserted
> > disk, and was timing out while trying to select the channel when forced
> > with an atacontrol reinit.
> 
> Do not use this for safety/raid1.
> 
> I have seen several reports that the NVidia software raid wipes out
> the intact drive on warm boot when the machine reboots with one drive
> bad.  Seems to do a random pick which drive to mirror to which other.

Thanks for confirming this - that is exactly the same behaviour as I
have witnessed.  I've even managed to lose data from the disks without
any OS involvement, purely by using the RAID BIOS.  I'm amazed how bad
this controller is.

> gmirror will do fine for raid1, and performs adequately (not as well
> as Linux software raid1, but probably better than NVidia anyway).

I've built one of the servers using native ataraid ("FreeBSD
PseudoRAID"), which seems to work fine (except for the lack of hot-swap,
which as I say, apparently doesn't work under any non-XP operating
system anyway).

Gavin
Received on Tue Nov 29 2005 - 12:37:26 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:48 UTC