Re: Lots of con-current I/O = resets SATA link? (2.6.25.10)

[Mattias Wadenstein]

On Sat, 5 Jul 2008, Justin Piszcz wrote:

> On Sat, 5 Jul 2008, Robert Hancock wrote:
>
> > Justin Piszcz wrote:
> > > > Can you post your dmesg from bootup with the controller/drive detection?
> >
> > So you've got 6 drives in the machine. Intel chipsets normally seem pretty 
> > robust with AHCI.
> >
> > Are you certain that your machine has enough power to run all those drives 
> > properly? We've seen in a number of cases that power fluctuations or noise 
> > can cause these kinds of errors.
>
> I have a 650watt PSU (nice antec one) and the power draw of the box is 
> ~148watts w/ veliciraptors, ~250 when fully load all 4 cores + all 12 disks 
> writing.  I have turned off the irqbalance daemon and I am going to see if 
> the problem re-occurs.

Looking at the sum wattage number is really misleading for this. You need 
to dig out the specs for how many amps it can provide on the different 
voltages (5 and 12 volts). In particular, many modern PSUs have several 
separate 12V rails, where one (or more, some have the 12V supply split 
into 3 or 4 parts!) is used for CPU and GFX card power and usually only 
one is available for disks.

You can also have plenty of 12V left but run out of 5V, or the other way 
around. I've spent quite some time trying to find a PSU that would handle 
18 disks without costing too much. The splitting of the 12V power into 
separate rails and a general lack of 5V compared to what the disks need 
according to their specs just made it difficult, and I ended up bonding 
two PSUs together (linking the ground together with some custom cabling) 
to get a stable machine again.

/Mattias Wadenstein
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/