> > pre-patch-2.0.30-2 solves the problem or not, but I did notice a couple
> > inode operations now guarded by semaphores in the patch. I'll be doing
> > my stress testing and debugging off of the pre-patch first, and then off
> > of 2.0.29.
> >
> > As always, any help or suggestions would be greatly appreciated.
>
> The solution I used to solve a memory corruption problem we started getting
> regularly on our newsserver was to replace the AHA2940 with a real scsi
> controller. End of problem, no glitches since.
>
I've had some AHA2940UWs in 1.2.13 linux boxen for over a year or so now,
with no failures. I still don't liek them all that much tho, especially
not since I got a hold of a DPT, the ones with the Knight Rider lights.
We had a Soalris x86 box that was crashing almost anytime backups were
run because of a bad revision in a Buslogic controller, we replaced it
with a DPT(and spent 15 minutes running it with the case off to watch the
great lightshow) and it's now been up 50 days, under considerable load
and disk access (it's www, database, and NFS service, hub for network and
backups).
A good SCSI HBA is of extreme importance to the life of any server, I've
had my share of bad experiences with Adaptec's(mostly in server machines,
they have always worked fine in desktop workstations for me), and
Buslogics. I've never used an NCR tho.
Is there support for DPTs under the recent Linux kernels?