Bonding problems 3Com 3C980 cards

I have installed a Suse 9.1 system with 2 3Com 3C980-TX cards.
Both cards are enslaved for bonding under bond0. The bonding mode
is fault-tolerant.
I can get no reaction from the NIC's and the system says everything
is OK. Looking into /proc/net/bonding/bond0 all looks wel (same as
for the Intel cards).

When I replace the 2 3Com NIC's by 2 Inter Pro 100's everything works
perfectly.
Anyone ever seens this before ?

Gtx,

Arjan Smit

Hi Arjan,

does /proc/net/bonding/bond0 reflects it when you take away the cable
from one of the two NICs (Link state of that slave switches to "fail") ?

This was a driver issue in my case, it did not support everything needed
to detect a link failure, therefore it always showed "Link state: OK" ...

Kind regards,
Heiko

Arjan wrote:

> I have installed a Suse 9.1 system with 2 3Com 3C980-TX cards.
> Both cards are enslaved for bonding under bond0. The bonding mode
> is fault-tolerant.
> I can get no reaction from the NIC's and the system says everything
> is OK. Looking into /proc/net/bonding/bond0 all looks wel (same as
> for the Intel cards).
>
> When I replace the 2 3Com NIC's by 2 Inter Pro 100's everything works
> perfectly.
> Anyone ever seens this before ?
>
> Gtx,
>
> Arjan Smit


--
Meinberg radio clocks: 25 years of accurate time worldwide

Meinberg Radio Clocks
www.meinberg.de

Stand alone ntp time servers and radio clocks based on GPS, DCF77 and
IRIG. Rackmount and desktop versions and PCI slot cards.

> I have installed a Suse 9.1 system with 2 3Com 3C980-TX cards.
> Both cards are enslaved for bonding under bond0. The bonding mode
> is fault-tolerant.
> I can get no reaction from the NIC's and the system says everything
> is OK. Looking into /proc/net/bonding/bond0 all looks wel (same as
> for the Intel cards).
>
> When I replace the 2 3Com NIC's by 2 Inter Pro 100's everything works
> perfectly.
> Anyone ever seens this before ?

You may need to specify the 'use_carrier' parameter. e.g.

alias bond0 bonding mode=1 miimon=100 primary=eth1 use_carrier=0

I tried the use_carrier=0 setting (also use_carrier=0) with no luck.
There might be more to it then this.
When I try to use ethtool to check settings for Intel cards it
returns te info like it should. When used for the 3COM
cards it says it can't get any info (No data available. is
printed on the terminal).



Bob Smith <(E-Mail Removed)> wrote in message news:<(E-Mail Removed)>...
> > I have installed a Suse 9.1 system with 2 3Com 3C980-TX cards.
> > Both cards are enslaved for bonding under bond0. The bonding mode
> > is fault-tolerant.
> > I can get no reaction from the NIC's and the system says everything
> > is OK. Looking into /proc/net/bonding/bond0 all looks wel (same as
> > for the Intel cards).
> >
> > When I replace the 2 3Com NIC's by 2 Inter Pro 100's everything works
> > perfectly.
> > Anyone ever seens this before ?
>
> You may need to specify the 'use_carrier' parameter. e.g.
>
> alias bond0 bonding mode=1 miimon=100 primary=eth1 use_carrier=0

Heiko Gerstung <heiko.gerstung_removeme_@meinberg.de> wrote in message news:<if6n42-(E-Mail Removed)>...
> Hi Arjan,
>
> does /proc/net/bonding/bond0 reflects it when you take away the cable
> from one of the two NICs (Link state of that slave switches to "fail") ?
>
> This was a driver issue in my case, it did not support everything needed
> to detect a link failure, therefore it always showed "Link state: OK" ...
>
> Kind regards,
> Heiko
>

When you wait a long time it will detect a fail (about 2 minutes!) and then
/proc/net/bonding/bond0 reflects the link failure correctly and passes over
communication to the other NIC.