Very Computer

Hi!

I have two computers: one running as server
(samba,nfs,apache,squid,leafnode,...) and router (ISDN) with Linux (SuSE
6.3) and the other one running Windows 98SE and Linux. The network
connection from windows to linux is perfect. Around 5MB via samba and a
few hundred KB to leafnode and when requesting a cached website from
squid. When running linux I got 1.5MB via NFS, about 4-8 KB via squid
(cached! website) and netscape even stalls when accessing leafnode.
I am desperate!!!! I have absolutely no idea where to problem can be...
I am thankful for every ideas, help, hints or even guesses.

Hardware of the workstation:
AMD K6-III/400
Asus P5A mainboard (ALI chipset - not recognized by linux)
Ram: 128 MB
nic: 3com 905B
grafic: elsa victory II
SB16 (not installed in linux)

Hardware of the server:
P100
Asus SP97XV (SiS chipset - recognized by linux)
Ram: 128 MB
nic: 3com 905B
grafic: Spea Mirage V7 (very old, no X installed)

Both running SuSE 6.3, kernel 2.2.13.

Thanks in advance, Martin

I was dismayed too when I realized that changing from Windows
to Linux on the desktop would mean inferior connection speed.
Samba on the fileserver does a much better job than NFS.
However things have been worse. Userland NFS in the 2.0.x kernel
was horrible compared to kernelspace 2.2.x.

If you have CPU time and memory to spare on the fileserver,
you _can_ increase the number of NFS daemons above the
guideline count. Double it and see what it does for your
performance.

If SuSE is anything like RedHat this involves changing the
value assigned to the script variable RPCMOUNTDCOUNT in
/etc/rc.d/init.d/nfs, then restarting the NFS service.

Svend

Are either of the Linux boxes reporting lots of errors or collisions in
"ifconfig"?

Have you tried running tcpdump during one of these snail-paced transfers?

-Tom
--
Tom Eastep               \    Opinions expressed here are

Shoreline, Washington USA  \    those of my employer
ICQ #60745924               \____________________________

Ok, now I found what: I got a LOT of net errors when running linux. If I
do an ifconfig eth0 just a few minutes after start up I get around 100
errors:
eth0      Link encap:Ethernet  HWaddr 00:10:4B:C5:18:CA
          inet addr:192.168.1.2  Bcast:192.168.1.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:3253 errors:91 dropped:0 overruns:0 frame:142
          TX packets:2601 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:100
          Interrupt:11 Base address:0xb800

The same on the server:
eth0      Link encap:Ethernet  HWaddr 00:10:4B:C5:18:9E
          inet addr:192.168.1.1  Bcast:192.168.1.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:382779 errors:2 dropped:0 overruns:0 frame:2
          TX packets:464589 errors:544 dropped:0 overruns:0 carrier:613
          collisions:0 txqueuelen:100
          Interrupt:10 Base address:0xb800

If I open one web site (which resided in squid's cache) running with
windows I doesn't get one additional error on the server. If I run linux
and open the same site, netscape stalls after a few seconds and I get
around 10-20 additional errors on the server.

WHY????????????????????????????????????????????????????????????????
As it works perfectly with windows, it can't be a hardware or cable
problem. So what is it?

Any help appreciated!
Martin

The client reports lots of RX-errors and the server lots of TX-errors. BUT
the server reports only errors, when connected to linux. If connected to
windows (on the same machine!) there isn't one single error (Doing in windows
and linux the same: looking at one website, which is cached by squid, with
netscape 4.7). So it can't be a hardware or cable problem. Any idea what I
could have misconfigured with my nic????

Quote:> Have you tried running tcpdump during one of these snail-paced transfers?

No, but I think the problem is the massive number of transmisson errors
reported by ifconfig. Correct?

Many thanks for your reply, Martin.

Problem solved: I just updated my nic driver and recompiled my kernel -
that's it!