Last week, I flashed one of my two systems from NJ to QQ_2. The system is
a dual 366/128MB(ECC) functioning as a Samba and web server for my 40GB+
MP3 collection, so the load is exceptionally light although network
traffic is heavy. The machine runs headless with a single 10/100 intel
NIC. It mounts two IDE Hard drives (IBM 27GBs) and a single IDE CDrom.
These are on the non-HPT controllers.
Prior to the upgrade, I was experiencing a freeze approximately every five
days. Believe me, fscking 54GB of drive space gets really old.
After flashing to QQ I went seven days without a freeze. During this
time, to stress-test the box, I ran overclocked to 550MHz. I also had two
SETI@Home procs going.
Next, I flashed my other machine, which is identical to the first except
it has a single 15GB drive and one CDRom (both machines were assembled at
the same time with parts from the same companies). This machine was not
crashing, but since it was only functioning as a primary nameserver for
about 80 domains and a caching nameserver for my home LAN and for
Geek.NET's systems, the traffic to it was exceptionally light. In the
past, however, I have had problems getting this machine to run SETI@Home
at 550MHz. The SETI process would die after about 15 minutes.
After flashing the second machine (ns1.geek.net) to QQ2, I was able to run
it for 72 hours at 550MHz without crashes and without SETI dying. On a
whim I upped the FSB to something above 100MHz and the machine didn't like
it; I had to discharge the CMOS to regain access to the console.
I am using RH6.0 with security patches and a 2.2.14 kernel on both
machines. I am _not_ using "noapic" on either machine. I believe that
the 2.2.x series doesn't display APIC warnings, so I don't know if I'm
having problems.
My conclusion is that QQ2 goes a long way towards fixing the problems
we've been seeing since the BP6 was introduced. However, I need at least
90 days of uptime (minus maintenance downtimes) before I'm willing to back
off of my unwillingness to deploy BP6-based (or even abit-based) systems.
All the voodoo we've been having to go through has left a bad taste in my
mouth.
I'm also rather disturbed at the lack of information coming out of abit.
I know they know this list exists, and it'd be really nice to have an abit
engineer here and contributing to the discussion if only to verify they
are addressing the problems.
In summary...
Don't complain about problems unless you are at QQ. :-)
Questions? Comments?
=-=-=-=-=-=
Robert Hayden [EMAIL PROTECTED] UIN: 16570192
--
=- To unsubscribe, email [EMAIL PROTECTED] with the -=
=- body of "unsubscribe linux-abit". -=
