Ethernet on 43P: ok on hub, not ok on switch

Ethernet on 43P: ok on hub, not ok on switch

Post by Anthony Dyso » Fri, 24 Dec 1999 04:00:00



I have a 43P running AIX 4.2.x, standard on-board ethernet, and a couple of
plug-and-play Netgear switches and hubs. The other hosts on the LAN are a
mixture of linux and Windows NT systems, with the entire LAN behind a Cisco
router. The ethernet adapter is configured to operate in half-duplex mode,
and I am using the en0 interface.

When I connect the 43P directly to a switch port, ping connections to and
from it work correctly. Also, I can telnet from the 43P to another host.
When trying to establish a telnet connection from another host to the 43P,
everything stops after the username and password are entered. Looking at
netstat -a on the 43P I see an established telnet connection, but with
around 1000 packets in the outbound queue. Similar behaviour is exhibited
with other TCP applications, e.g. ftp and http.

When I connect the 43P to a hub* off the same switch everything
functions perfectly. I've tried this with a couple of different hubs, with
no problems.

Wondering whether the problem could be related to ethernet frame standards,
I tried switching over to et0. Using et0 I am able to ping the Cisco, but
get responses from no other host in the LAN. Pinging hosts on the same hub
was also unsuccessful.

I also tried replacing the cable(!), and tried various combinations of half
and full duplex.

Can anyone suggest what might be causing this problem, or what further tests
I could try?

Thank you,
    Tony Dyson

--
=================================================================
Dr. Anthony Dyson                       Ph:  +41-61-9272077
Triangle Micro Research AG              Fax: +41-61-9220372
Zeughausplatz 11

=================================================================

This signature has been specially enhanced for viewing
with a monospaced font.

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Snogfest Hosebeas » Sat, 25 Dec 1999 04:00:00




Quote:> Can anyone suggest what might be causing this problem, or what
further tests
> I could try?

> Thank you,
>     Tony Dyson

check "netstat -m"
also
"lsattr -El ent0"
--
R1, TDM.Ghairly Ghast Ghat Ghuq.
Sometime AIX geek, and below-average volleyballer.
Youth and fitness will always lose to experience and treachery

Sent via Deja.com http://www.deja.com/
Before you buy.

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by <ye.. » Sat, 25 Dec 1999 04:00:00


I prefer to set the ent0 dev speed manually.  ie 100Mbps half-dup
or whatever.  I normally don't like setting it to Auto-Neg.
--

The opinions expressed are my own and do not reflect those of IBM.
 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Timothy J. Boga » Tue, 28 Dec 1999 04:00:00



<snip>

>Can anyone suggest what might be causing this problem, or what further tests
>I could try?

>Thank you,
>    Tony Dyson

>--
>=================================================================
>Dr. Anthony Dyson                       Ph:  +41-61-9272077
>Triangle Micro Research AG              Fax: +41-61-9220372
>Zeughausplatz 11

>=================================================================

<snip>

Well, some time ago, the project I was on was cancelled before we could
get the vendors to sort things out...but as I recall, it did in fact
boil down to the lack of true standardization on auto-negotiating.  
Workaround was to manually set all machine and hub interface speeds.

Sigh.  Still beats the performance of the old BNC stuff....8-)

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Anthony Dyso » Wed, 29 Dec 1999 04:00:00


For those who are interested, the workaround for this problem turns out to
be reducing the MTU for the en0 interface driver.

Pinging with progressively larger packets whilst attached to the switch
showed that as of about 768 bytes the link was unreliable, and as of about
1000 bytes it did not function at all. Attached to the hub, ping packets of
size 4096 were being returned with 100% reliability, and I did not test
further. Reducing the MTU to 600 has allowed me to connect the 43P to the
switch with no further problems.

    Tony Dyson

--
=================================================================
Dr. Anthony Dyson                       Ph:  +41-61-9272077
Triangle Micro Research AG              Fax: +41-61-9220372
Zeughausplatz 11

=================================================================

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Nicolas DE METZ-NOBLA » Tue, 04 Jan 2000 04:00:00


I have seen this problem with a 7248-120 (old 43P) and a CISCO Catalyst
2820 switch: I had to put a small intermediate ethernet hub and
everything works fine. This definitively looks like a bug on the 7248
side, because I had absolutly no problems to connect any other systems
including RS/6000 7009 (C20) using an ethernet transceiver.


> I have a 43P running AIX 4.2.x, standard on-board ethernet, and a couple of
> plug-and-play Netgear switches and hubs. The other hosts on the LAN are a
> mixture of linux and Windows NT systems, with the entire LAN behind a Cisco
> router. The ethernet adapter is configured to operate in half-duplex mode,
> and I am using the en0 interface.

> When I connect the 43P directly to a switch port, ping connections to and
> from it work correctly. Also, I can telnet from the 43P to another host.
> When trying to establish a telnet connection from another host to the 43P,
> everything stops after the username and password are entered. Looking at
> netstat -a on the 43P I see an established telnet connection, but with
> around 1000 packets in the outbound queue. Similar behaviour is exhibited
> with other TCP applications, e.g. ftp and http.

> When I connect the 43P to a hub* off the same switch everything
> functions perfectly. I've tried this with a couple of different hubs, with
> no problems.

> Wondering whether the problem could be related to ethernet frame standards,
> I tried switching over to et0. Using et0 I am able to ping the Cisco, but
> get responses from no other host in the LAN. Pinging hosts on the same hub
> was also unsuccessful.

> I also tried replacing the cable(!), and tried various combinations of half
> and full duplex.

> Can anyone suggest what might be causing this problem, or what further tests
> I could try?

> Thank you,
>     Tony Dyson

> --
> =================================================================
> Dr. Anthony Dyson                       Ph:  +41-61-9272077
> Triangle Micro Research AG              Fax: +41-61-9220372
> Zeughausplatz 11

> =================================================================

> This signature has been specially enhanced for viewing
> with a monospaced font.

--
Nicolas de METZ-NOBLAT   Phone: +41 22 76 72023 (direct)
CERN - PS division              +41 79 201 3070
CH-1211 Geneva 23        Fax:   +41 22 76 79145

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Gerhard Gonte » Tue, 04 Jan 2000 04:00:00



> I have seen this problem with a 7248-120 (old 43P) and a CISCO Catalyst
> 2820 switch: I had to put a small intermediate ethernet hub and
> everything works fine. This definitively looks like a bug on the 7248
> side, because I had absolutly no problems to connect any other systems
> including RS/6000 7009 (C20) using an ethernet transceiver.

RS 6000's with 10/100 MBit ethernet and CISCO catalysts apparently
give each other a hard time for speed autonegotiation.  A solution is
to set the ports to 100 MBit full duplex on both sides (RS 6000 and
the catalyst).  The C20 has a 10 MBit ethernet interface and does not
have a problem with autonegotiation.

+gg

 
 
 

Ethernet on 43P: ok on hub, not ok on switch

Post by Kimo Bonnelyck » Sat, 15 Jan 2000 04:00:00


I have also seen this problem when using NIM and token ring.  It seems (for
us), that
unless there is a hub between the switch and the machine(the machine being
loaded), we get the same
problems.  This was fixed with a new switch.  We were not really sure what
caused the problem, but
have a couple of ideas, one being a speed negotiation problem between 4/16mb
for token ring.  Another
being a packet issue (NIM not sending the correct number of packets to open
and close the swtich, kinda of
hard to explain).

Regards

Kimo Bonnelycke
IBM AIX Software Distribution



> > I have seen this problem with a 7248-120 (old 43P) and a CISCO Catalyst
> > 2820 switch: I had to put a small intermediate ethernet hub and
> > everything works fine. This definitively looks like a bug on the 7248
> > side, because I had absolutly no problems to connect any other systems
> > including RS/6000 7009 (C20) using an ethernet transceiver.

> RS 6000's with 10/100 MBit ethernet and CISCO catalysts apparently
> give each other a hard time for speed autonegotiation.  A solution is
> to set the ports to 100 MBit full duplex on both sides (RS 6000 and
> the catalyst).  The C20 has a 10 MBit ethernet interface and does not
> have a problem with autonegotiation.
> I have a 43P running AIX 4.2.x, standard on-board ethernet, and a couple of
> plug-and-play Netgear switches and hubs. The other hosts on the LAN are a
> mixture of linux and Windows NT systems, with the entire LAN behind a Cisco
> router. The ethernet adapter is configured to operate in half-duplex mode,
> and I am using the en0 interface.

> When I connect the 43P directly to a switch port, ping connections to and
> from it work correctly. Also, I can telnet from the 43P to another host.
> When trying to establish a telnet connection from another host to the 43P,
> everything stops after the username and password are entered. Looking at
> netstat -a on the 43P I see an established telnet connection, but with
> around 1000 packets in the outbound queue. Similar behaviour is exhibited
> with other TCP applications, e.g. ftp and http.

> When I connect the 43P to a hub* off the same switch everything
> functions perfectly. I've tried this with a couple of different hubs, with
> no problems.

> Wondering whether the problem could be related to ethernet frame standards,
> I tried switching over to et0. Using et0 I am able to ping the Cisco, but
> get responses from no other host in the LAN. Pinging hosts on the same hub
> was also unsuccessful.

> I also tried replacing the cable(!), and tried various combinations of half
> and full duplex.

> Can anyone suggest what might be causing this problem, or what further
tests
> I could try?

> Thank you,
>     Tony Dyson

> +gg

 
 
 

1. Printcap OK, Postscript OK, together NOT OK - HELP!

I am trying to setup Postscript support for printing with lpr on a Laserjet 4L.
I have modified printcap and written a filter using ghostscript, but not
success.  Can somebody help?  Here is my setup:
printcap:       lpps|hpljet3:\
                        :lp=/dev/lp1:\
                        :sd=/usr/spool/lp1:\
                        :sh:\
                        :sf:\
                        :mx#0:\
                        :of=/usr/spool/lp1/hpps:
hpps:           gs -q -sDEVICE=ljet3 -sOutputFile=- -

When I use "cat file.ps | hpps | lpr" (with my text-only lp setup) it works ok,
but when I try to use "lpr -Plpps file.ps" I get "Ghostscript: Unrecoverable
error, exit code 1".  Any ideas?

bruno

2. nfsd and exporting /lib

3. "nu bee:-)" ok ok ok

4. Buying off Qt

5. Modem Problems: dial-out is ok/dial-in is not ok (Solaris 2.6)

6. route on DU4.0B

7. Mounting Nextstep Partitions

8. Ping ok, netscape, gftp, lynx not ok

9. AWE32 NOT DETECTED PNP OK PNP OK

10. 2x IDE-HD ok+1x Mits.4x CDROM NOT ok!

11. 2.1.35 /PNP OK/SB OK but AWE32 not

12. serial console: LILO+getty ok, kernel not ok