Filesystem corruption when drives are mirrored, but not otherwise

[Norberto Meijome]

On Thu, 11 May 2006 21:59:49 +1000
Alastair Rankine <arsptr at optusnet.com.au> wrote:

> On 11/05/2006, at 10:59 AM, Norberto Meijome wrote:
> 
> > On Wed, 10 May 2006 22:45:10 +1000
> > Alastair Rankine <arsptr at optusnet.com.au> wrote:
> >
> >>  Disabling write
> >> caching seems to have fixed the problem.
> >
> > excuse my ignorance, but where do you do this? BIOS?
> 
> Set hw.ata.wc="0" in /boot/loader.conf
> 
> See tuning(7) for more info.

thx should have thought of this :) 

> 
> > I'm running GEOM mirrors on amd64 with SATA (I assume you mean PATA  
> > when you
> > say ATA).
> 
> Nope, I'm using SATA on a Promise FastTrak S150 TX2plus.
> 
> What controller are you using?

Software raid actually - the box has Nvidia RAID in the bios but didnt work (at
all?) so ended up using gmirror in different layouts. A low end box :(

The controllers are 2 x nVidia nForce CK804 SATA300.

 atapci1: <nVidia nForce CK804 SATA300 controller> port
0x1440-0x1447,0x1434-0x1437,0x1438-0x143f,0x1430-0x1433,0x1410-0x141f mem
0xdd002000-0xdd002fff irq 22 at device 7.0 on pci0

& 

atapci2: <nVidia nForce CK804 SATA300 controller> port
0x1458-0x145f,0x144c-0x144f,0x1450-0x1457,0x1448-0x144b,0x1420-0x142f mem
0xdd003000-0xdd003fff irq 23 at device 8.0 on pci0

Drives are 4 x WDC WD2500JS-00MHB0 02.01C03


> 
> Do you get a decent write speed?
> 
> cstream -o tmpfile -v 1 -n 384m -i -
> 
> ... is a good benchmark for sequential writes (cstream is in the ports).

[numard at cerberus] [Fri May 12 01:56:24 2006]
/tmp
$ cstream -o tmpfile -v 1 -n 384m -i -
402653184 B 384.0 MB 6.91 s 58254333 B/s 55.56 MB/s

definitely not great throughput, but it'll have to do for now :-(

thanks a lot for the time to share your knowledge :)

Beto