Hi Managers,
Another big thank you to all the replies. I would have summarized
sooner but I needed time to monitor and implement the changes.
This problem was due to a miss - match in the duplex setting. I
corrected the problem by manually setting the speed and duplex
type for the interface at the console level:
>>set ewa0_mode fastFD
>>init
There is a typo in the console manual about this. It says this is for
SCSI and reads fast duplex, twisted-pair should be the value. I
never tried it as Selden Ball pointed it out before I attempted it.
Once this was done I manually set the switch to 100/full duplex for
the particular port the server was on. The switch is registering
100/full duplex and there hasn't been anything like the same amount
of collisions.
Thanx to:
Selden Ball
Jo Fletcher
Oisin McGuinness
Udo de Boer
Brain H
Jerry Hummel
David Warren
Dan Riley ~ I thought that amount of collisions was a problem
personally.
Dp.
-----------------Original message -------
Hi Managers,
I hope this isn't off topic. We have 2 Alphas; First, an old Dec Alpha
3000/400. I am not sure what model NIC it has installed. It it runs at
10MBps and is connected using Cat5 utp to a DEC repeater 90L+.
The
second a new Compaq DS10 which has an on-board 10/100MB NIC.
Again
this has a utp connection but connected to a Dlink 10/100 switch
(model DES 2334).
Below I have copied the netstats for both servers, please excuse the
formatting. The old server has a database, user files, and lots of
compressed data files. It is an old work horse and seems to be doing
fine the newer. Compaq shows a fantastic amount of network
collisions
for a machine that isn't really live although I have been dumping lots
of hi-res compressed images files onto it. The ration is about 60:40.
If you look at the figures the old server has sent out 11 million or
so packets and only had 20975 collisions. The Compaq has
sent/received
5727357/4184668 packets and had 2367528 collisions. To my mind
this is
worrying.
I have not recompiled the kernel on the new server since I took
stock. The DEC Alpha has been up for one day, the compaq for 5
days.
Does anyone have any thoughts on these collisions? The man page
on netstat does say a great deal. I am not sure whether this is a
server problem or its on the LAN. Thanx Dp.
---------------- DEC server ----------------------------
Name Mtu Network Address Ipkts Ierrs Opkts
Oerrs Coll ln0 1500 <Link> 08:00:2b:3b:74:65 11004295
0 11547587 0 20975 ln0 1500 DLI none
11004295 0 11547587 0 20975 ln0 1500 194.200.237.128
helios
11004295 0 11547587
0 20975
sl0* 296 <Link> 0 0 0
0 0 lo0 4096 <Link> 38687
0 38687 0 0 lo0 4096 loop localhost
38687 0 38687 0
0
ppp0* 1500 <Link> 0 0 0
0 0
--------------- Compaq server -------------------
Name Mtu Network Address Ipkts Ierrs Opkts
Oerrs Coll lo0 4096 <Link> Link#4 98040
0 98040 0
0
lo0 4096 loop localhost 98040 0 98040
0
0
sl0* 296 <Link> Link#3 0 0 0
0 0 tu0 1500 <Link> 0:10:64:30:5a:61 5727357
0 4184668 0 2367528 tu0 1500 DLI none
5727357 0 4184668 0 2367528 tu0 1500 194.200.237.128
nemesis 5727357 0 4184668 0 2367528 tu1 1500
<Link> 0:10:64:30:5a:5f 0 0 10101 10101 0
tu1 1500 DLI none 0 0 10101
10101 0 tun0* 1280 <Link> Link#5 0
0 0 0 0
------- End of forwarded message -------
~~
Dermot Paikkos * dermot_at_sciencephoto.com
Network Administrator _at_ Science Photo Library
Phone: 0207 432 1100 * Fax: 0207 286 8668
Received on Wed Jul 11 2001 - 15:20:11 NZST