Original question below.
Thanks to all those who provided help!
The trick is not use auto negotiation. Instead explicitly
set the card to the speed you want to use. In console mode:
set ewa0_mode fastfd (for 100 full duplex)
or
set ewa0_mode fast (for 100 half duplex)
(this can apparantly also be set using the lan_config
command but I haven't played with this)
Along with one of those that emailed help I found that
fd wasn't completely reliable (the network seemed to stall
every 10 seconds or so for a second or two). I then tried
half duplex and all seems to be functioning well.
However.... I still have some mighty weird NFS performance
problems on the DS20, but no time this week to investigate this
further.
- Michael
----------------------------------------------------------------------
(I sent this to comp.unix.osf.osf1, but it may be more
appropriate here)
I've just turned auto-negotiation on on the ethernet interface on
our DS20 (OSF1 kandinsky V4.0 1229 alpha):
Jun 27 15:03:15 kandinsky vmunix: tu0: DECchip 21140: Revision:
2.2
Jun 27 15:03:15 kandinsky vmunix: tu0: auto negotiation capable
device
Jun 27 15:03:15 kandinsky vmunix: tu0 at pci2001 slot 2
Jun 27 15:03:15 kandinsky vmunix: tu0: DEC TULIP (10/100)
Ethernet Interface, hardware address: 00-06-2B-01-29-A1
Jun 27 15:03:15 kandinsky vmunix: tu0: auto negotiation on: will
advertise 100BaseTX (UTP) port: full duplex
previously, it seemed that by default auto-negotiation was off
aand the default settings of the card meant that it ran at
10MBits/s.
This change fixed this sort of. Now the card runs at 100MBits/s
but periodically resets itself (every minute or so):
Jun 27 17:37:26 kandinsky vmunix: tu0: link down
Jun 27 17:37:30 kandinsky vmunix: tu0: link up: negotiated
100BaseTX: full duplex
Jun 27 17:38:32 kandinsky vmunix: tu0: link down
Jun 27 17:38:35 kandinsky vmunix: tu0: link up: negotiated
100BaseTX: full duplex
Jun 27 17:39:53 kandinsky vmunix: tu0: link down
Jun 27 17:39:57 kandinsky vmunix: tu0: link up: negotiated
100BaseTX: full duplex
Jun 27 17:44:17 kandinsky vmunix: tu0: link down
Jun 27 17:44:20 kandinsky vmunix: tu0: link up: negotiated
100BaseTX: full duplex
As you can see there is nothing regular in the way this error
occurs but in gerneral the link usually stays up for around
1 to 1 1/2 minutes before going down. I've played around with
cables and ports but to no avail.
Anybody have a clue as to what may be going wrong? All our
other boxes are correctly coperating at 100MB/s (including our
ES40 with a similar NIC).
Thanks,
- Michael
Received on Fri Jun 30 2000 - 12:20:29 NZST