RE: How much space needed for a web server install?

[Scott Stone <SStone@example.com>]

Hmm, I see:

> For this one
> executes the following batch file under Windows, whereby source and target
> path should be on different drives and the file should be more than 
> 100 MBytes large. 

This indicates to me that it's a problem transferring BETWEEN drives, and
not a problem with a SINGLE drive.  So I might have problems installing
stuff from my CDROM, but with a single disk it should work, if the above is
to be believed... unfortunately the patch that is referenced in the
forwarded message below has a specific caveat that it does nothing on the
AMD-761 Northbridge boards.. :(

But, it also says:

> On Main boards with the AMD Northbridge AMD-761 and the VIA Southbridge
686B the error
> could theoretically likewise emerge, but at present the Patch does
> nothing on these systems.

indicating that it might NOT be a problem on the AMD-761, if I'm reading
this right?  I've sent out emails to VIA, Gigabyte, and the vendor from
where I bought the board asking for more information on this..

but my box has been running for 2 weeks with no problems...

-----------------------------------------------------
Scott M. Stone <sstone@example.com>
Senior Technical Consultant - UNIX and Networking
Taos, the Sysadmin Company - Santa Clara, CA


-----Original Message-----
From: Tobias Diedrich [mailto:ranma@example.com]
Sent: Friday, May 11, 2001 3:52 PM
To: 'tlug@example.com'
Subject: Re: How much space needed for a web server install?


Scott Stone wrote:

> well, so far, so good.  Have you found any specific references to this
> happening with the AMD761/VIA686B bridge?

That's the "Even with the AMD-761 Northbridge" part...

> and you're sure it's the 686B that's the problem?  All the athlon DDR
boards
> seem to use that southbridge from what I have seen... 

I just looked at the magazines newsticker and found this (from May 10th):
[Umm, translated with babelfish from german to english and improved a bit by
me]

---------------------------------------------------------------------------
VIA publishes Patch against 686B-Bug 

The chip manufacturer VIA developed a Patch, which is to recover the error
in the VIA Southbridge. At present the Patch is in a beta phase, but the
hardware Site ViaHardware provides a prerelease Version for Windows
9x/ME/2000 for download. 

A first test in the c't laboratory ran positively: A particularly
susceptible
Mainboard is running since over 15 hours without errors after applying
the patch. If the board was equipped with Creative soundcard, when
transferring large quantities of data over the IDE channels of the
Southbridge
errors had occurred, original and copy were different. All measures and also
an BIOS update of the board manufacturer changed only the frequency of the
disturbances, but did not let them disappear. Only after removing the sound
card no more errors had occurred. With this board the Patch of VIA finally 
helped: Even with the old BIOS version and "optimized settings" all
datatransfers so far ran without disturbances. 

On closer inspection it shows up that the Patch changes the dispatching
procedure for the PCI bus: It increases the priority of the CPU ACCESSES in
relation to the bus master accesses. We could not determine a considerable
speed loss by the modified dispatching. 

In the present version the Patch can be installed on any Main board.
However it modifies the Northbridge register only with the VIA Chipsets
KX133, KT133 or KT133A and also only, if a Soundblaster Live! is in the
system. 
On Main boards with the AMD Northbridge AMD-761 and the VIA Southbridge 686B
the error could theoretically likewise emerge, but at present the Patch does
nothing on these systems.

[...]

Thus for the time being VIA remains at its statement that the error emerges
only in connection with Creative's Soundblaster Live. 

Diagnosis of the VIA bug remains difficult, since with system crashes the
guilty component is rarely clear. In the c't laboratory a test method
turned out as quite reliable, with which the system does not fall after a
data error , but continues to work with an error message. For this one
executes the following batch file under Windows, whereby source and target
path should be on different drives and the file should be more than 
100 MBytes large. 

set quelle=f:\irgendwo
set ziel=g:\woanders
set datei=\grossedatei
del test.txti
@example.com off
:anfang
copy /B %quelle%%datei% %ziel%
fc / B %quelle%%datei% %ziel%%datei% > test.one
type test.one
type test.one >> test.txt
goto anfang

In the c't laboratory no further activity on the PCI bus was necessary.
The presence of the sound card with installed drivers was already
sufficient,
in order to provoke the error.
( jow / c't) 
----------------------------------------------------------------------------

> Just trying to gauge to what degree I should be entering "oh !@#$" mode at
> this point :)

If they really have a fix for the Problem you just have to hope it will
be integrated into the linux kernel fast ^_^

But it might have ended up like the Intel MCH2 desaster...

-- 
Tobias
nya~ni ?

-----------------------------------------------------------------------
Next Technical Meeting:  Sat, May 12 13:30- 
Next Nomikai Meeting:    Fri, June (TBA) 19:30- Tengu Tokyo Eki Mae
-----------------------------------------------------------------------
more info: http://www.tlug.gr.jp           Sponsor: Global Online Japan