[syslinux] Memory upgrade breaks net boot (PXE + NFS Root)

[Alex Still]

On 2/10/07, Geert Stappers <stappers at stappers.nl> wrote:
>
> Op 09-02-2007 om 14:18 schreef H. Peter Anvin:
> > Curtis Doty wrote:
> > >
> > > Unfortunately, I've had similar results where memtest86+ failed to
> detect
> > > any errors, but a specific memory stick was at fault. I'm only
> guessing it
> > > was a combination of memory access with rapidly firing interrupts from
> the
> > > network card or backplane during boot. In the end, the only way to
> > > identify faulty DIMM was good ol' trial-and-error elimination testing.
> > >
> >
> > Given that this is happening over a range of machines I would guess
> > memory bus overloading is probably the most likely cause.  Downclocking
> > the memory might be an interesting experiment.
>
> I do hope that is possible.
> Also I wonder what the impact on performance will be.
>
>
>
> And then is there the NIC fighting for the memory bus.
>
> My advice is to contact the hardware vendor about the issue.
> He might confirm that the NIC indeed uses main memory.



Haven't had time to look further into it this week, but now we're in touch
with IBM about this issue.
Something I didn't mention (don't think it matters but well..) is that the
new 4g of RAM are ECC, and the original 2g weren't.
It's definitly the memory change that breaks things : we have a total of 84
machines, only the half that has been upgraded has these problems.
I'll post a summary of what we find, if anything !
Thanks for your comments and insight,

-- 
Alex