[PLUG] SATA II, mdadm and soft power off

[John Jason Jordan]

On Sat, 12 May 2007 13:04:08 -0700
"Larry Brigman" <larry.brigman at gmail.com> dijo:

> > I thought of that, but I'm not sure if the PS units they have would
> > have the proper connectors for this motherboard, plus it needs 12 volts
> > separately for the CPU with a funny four-pin square connector I've
> > never seen before.
> >
> > As for the disk operation, I'm currently running mhdd from the Ultimate
> > Boot CD. No errors so far, but we'll see how it goes. I think Galen is
> > right and I need to run something that tests both disks at the same
> > time to see if the problem is in mdadm or in the controller. So far it
> > appears that there is no problem with the power running either disk by
> > itself.

> Is the power for the drives thru the motherboard or direct from the PSU?

Directly from the PSU. And since the new Antec PSU has two 12 volt
rails it has two SATA power connectors, each with two plugs on it. I
connected each drive to a different one, presuming that this means each
one is on a different rail. 

As for mhdd, it finished. The first disk ran nicely until it got to
around 80% with 99.99% of the blocks <3ms, then  suddenly the number of
blocks that were >3 and <10ms increased dramatically. At the end there
were three blocks that were >10 and <50ms. The average speed was about
80,000 kb/s. The second drive also did fine until about 80%, then the
number of slower blocks started increasing dramatically. At the end it
had 759 >10ms and <50ms, and the average speed was only 64,800 kb/s.
However, mhdd did not announce any errors as such. That is, it may have
been slower on one drive than the other, but evidently it finally read
all the blocks correctly. 

When it finished the second drive I decided to do a scan with the remap
switch on, starting with the second, slow drive. When it finished it
gave me more or less the same results as when I ran it without the remap
switch. I've scoured the scant documentation and forum for mhdd, but I
can't find anything that tells me how slow a block has to be before
mhdd will remap it, or if there is a way to change the threshhold.
Evidently 50 ms doesn't trigger a remap.

What bothers me about these results is that the second drive is quite a
bit slower than the first one. And timing was one of the things iozone
was bitching about. 

Quoting Dale:
 >I'll have to agree
> with Galen here, I think it's time to look for a new motherboard
> -- one without nVidia chipsets.  (Not to be confused with their
> graphics chips!)  It's possible you could have gotten two bad
> power supplies in a row (hey, it's happened to me), but it's
> really unlikely.

Of course, a sane person would just get a new motherboard. I paid $70
for the Abit and the Asus I found online is $84 at Fry's. If I bill out
at $40 an hour (I'm cheap), the 10+ hours I've spent on this so far is
ridiculous. But see, that isn't any fun. It's much cooler to be able to
sleuth down an obscure problem. It's like a giant sudoku puzzle, with
teams. On the one side there is Abit, nVidia, Western Digital, and a
host of others. On our side we have cool smart people. Of course, we
also have Linux on our side, so ultimately we will win. But that
doesn't mean it won't be an exciting adventure with clouds of bits
a-flying. And the frosting on the cake is the extra knowledge I will
gain from the experience. 

But I do have my limits. At some point I will pull the plug and go get
a different motherboard. :)