Bonding - slave nic keeps swapping
Hello list,
I'm running a couple of old Sun Netra T1 105 boxes which has 2 nic's:
eth0: HAPPY MEAL (PCI/CheerIO) 10/100BaseT Ethernet 08:00:20:c2:33:f8
eth1: HAPPY MEAL (PCI/CheerIO) 10/100BaseT Ethernet 08:00:20:c2:33:f8
I'm running Debian Sarge on sparc with the stable debian stock kernel
2.6.8-2-sparc64.
The purpose is to set up bonding so that if for some reason one of the nic's
should die or a cable gets pulled that everything keeps working.
It looks like my NIC's don't support mii stuff so I had to use the arp
option.
With the mii option I get the warnings that my nic's don't support link
detection, and indeed if I pull a cable nothing is detected.
So the bonding module gets loaded with these options:
options bonding mode=1 arp_interval=1000 arp_ip_targets=xx.xx.72.65
and in the interfaces file I got this:
auto bond0
iface bond0 inet static
address xx.xx.72.70
netmask 255.255.255.192
network xx.xx.72.64
gateway xx.xx.72.124
pre-up ifconfig eth0 up
pre-up ifconfig eth1 up
## up /sbin/ifenslave bond0 eth0 eth1
down /sbin/ifenslave -d bond0 eth0 eth1
up /sbin/ifenslave bond0 eth0
up /sbin/ifenslave bond0 eth1
post-down ifconfig eth0 down
post-down ifconfig eth1 down
Everything comes up nicely but with mode set to 1 it keeps swapping nics:
---------------
Jan 6 16:30:30 aberlour kernel: Ethernet Channel Bonding Driver: v2.6.0
(January 14, 2004)
Jan 6 16:30:30 aberlour kernel: bonding: ARP monitoring set to 1000 ms with
1 target(s): xx.xx.72.65
Jan 6 16:31:02 aberlour kernel: bonding: Warning : no link monitoring
support for eth0
Jan 6 16:31:03 aberlour kernel: bonding: bond0: making interface eth0 the
new active one 0 ms earlier.
Jan 6 16:31:03 aberlour kernel: bonding: bond0: enslaving eth0 as an active
interface with an up link.
Jan 6 16:31:03 aberlour kernel: bonding: bond0: link status down for active
interface eth0, disabling it
Jan 6 16:31:03 aberlour kernel: bonding: Warning : no link monitoring
support for eth1
Jan 6 16:31:03 aberlour kernel: bonding: bond0: making interface eth1 the
new active one 0 ms earlier.
Jan 6 16:31:03 aberlour kernel: bonding: bond0: enslaving eth1 as an active
interface with an up link.
Jan 6 16:31:03 aberlour kernel: bonding: bond0: link status down for active
interface eth1, disabling it
Jan 6 16:31:03 aberlour kernel: bonding: bond0: making interface eth0 the
new active one 0 ms earlier.
Jan 6 16:31:03 aberlour kernel: bonding: bond0: link status down for active
interface eth0, disabling it
Jan 6 16:31:06 aberlour kernel: eth0: Link is up using external transceiver
at 100Mb/s, Full Duplex.
Jan 6 16:31:06 aberlour kernel: eth1: Link is up using external transceiver
at 100Mb/s, Full Duplex.
Jan 6 16:31:08 aberlour kernel: bonding: bond0: making interface eth0 the
new active one.
Jan 6 16:31:08 aberlour kernel: bonding: bond0: eth0 is up and now the
active interface
Jan 6 16:31:08 aberlour kernel: bonding: bond0: link status down for active
interface eth0, disabling it
Jan 6 16:31:08 aberlour kernel: bonding: bond0: making interface eth0 the
new active one.
Jan 6 16:31:08 aberlour kernel: bonding: bond0: eth0 is up and now the
active interface
Jan 6 16:31:08 aberlour kernel: bonding: bond0: link status down for active
interface eth0, disabling it
Jan 6 16:31:08 aberlour kernel: bonding: bond0: making interface eth0 the
new active one.
Jan 6 16:31:08 aberlour kernel: bonding: bond0: eth0 is up and now the
active interface
And so on, every second.
However if I use mode = 0 (round robin) everything seems to work, if I pull
one of the cables only half the packets arrive until I plug it in again.
Mode = 3 (broadcast) works fine too, every packet arrives if I pull a cable.
But from articles I saw that mode = 1 (with a spare slave) is the most
common one.
Can someone tell me what's going on, why does it keep thinking the interface
is down and swaps slaves endlessly?
Thanks for your reply !
Kind regards,
Nick De Decker
Reply to: