topic Re: Bonding Problem in Operating System - Linux

Bonding Problem

Alvin_14 — Sun, 02 Dec 2007 05:19:56 GMT

I've got 4 interfaces, which i've bonded into 2.

bond0 - eth0, eth2 - 10.1.1.1
bon1 - eth1, eth3 - 10.1.1.2

First problem:

while bond0 is up i do ifdown bond1
- i can still ping bond0 which is good

but while bond1 is up and i do ifdown bond0
- there is delay when i ping bond1.. it takes a few minutes before i can ping.

Second problem:

while bond0 is up with both interfaces up, i do ifconfig eth2 down..
i can still ping bond0 and bond1 which is good.

while bond0 is up with both interfaces up, i do ifconig eth0 down..
like the first problem it takes a few seconds before i can ping both bond0 and bond1.

Third problem:
while bond0 is up with both interfaces up, i do ifconfig eth0 down and ifconfig eth2 down.

now i cannot ping bond1. it seems that bond0 is still logically alive thus the routing table is not updated.

Re: Bonding Problem

Rob Leadbeater — Sun, 02 Dec 2007 10:49:42 GMT

Hi,

From where are you pinging ?

I've never used bonded interfaces, but it sounds like you're hitting problems because of having two interfaces on the same network. What gateways/routes are set up ?

What subnet masks are you using on each interface ?

Cheers,

Rob

Re: Bonding Problem

Alvin_14 — Mon, 03 Dec 2007 09:17:23 GMT

from another server on the same subnet... for example, 10.1.1.3.. yes the problem seems to be becoz of the same subnet on the same machine.. but it's the application requirement.

Re: Bonding Problem

Alvin_14 — Mon, 03 Dec 2007 09:18:18 GMT

255.255.255.128

Re: Bonding Problem

Matti_Kurkela — Mon, 03 Dec 2007 09:25:38 GMT

You can use IP aliases on bonded interfaces too. Your second IP address might be configured like this:

ifconfig bond0:1 10.1.1.2 netmask 255.x.x.x

Now the application can use "bond0:1" just as if it were a "real" network interface, but the OS's network layer won't have problems related to having multiple real interfaces on the same subnet.

MK

Re: Bonding Problem

Steven E. Protter — Mon, 03 Dec 2007 13:52:47 GMT

Shalom,

Could be an error in the /etc/modprobe.conf configuration

Also could be a proflem in the ifcfg files in /etc/sysconfig/network-scripts

There may be a typo causing a conflict.

SEP

Re: Bonding Problem

Alvin_14 — Tue, 04 Dec 2007 01:37:56 GMT

here's my modprobe.conf:

alias bond0 bonding
options bond0 primary=eth1 miimon=100 mode=1 max_bonds=4
alias bond1 bonding
options bond1 primary=eth4 miimon=100 mode=1 max_bonds=4
alias bond2 bonding
options bond2 primary=eth3 miimon=100 mode=1 max_bonds=4

quite weird though that it shows primary slave as eth0 even if eth0 is primary of bond0 and not bond1.. or is this behavior correct?

[/proc/net/bonding]more bond1
Ethernet Channel Bonding Driver: v2.6.1 (October 29, 2004)

Bonding Mode: fault-tolerance (active-backup)
Primary Slave: eth0
Currently Active Slave: eth1
MII Status: up
MII Polling Interval (ms): 100
Up Delay (ms): 0
Down Delay (ms): 0

Slave Interface: eth1
MII Status: up
Link Failure Count: 1
Permanent HW addr: 00:11:0a:62:b9:2d

Slave Interface: eth3
MII Status: up
Link Failure Count: 1
Permanent HW addr: 00:11:0a:63:6b:8f
[/proc/net/bonding]more bond2
Ethernet Channel Bonding Driver: v2.6.1 (October 29, 2004)

Bonding Mode: fault-tolerance (active-backup)
Primary Slave: eth0
Currently Active Slave: None
MII Status: down
MII Polling Interval (ms): 100
Up Delay (ms): 0
Down Delay (ms): 0
[/proc/net/bonding]more bond3
Ethernet Channel Bonding Driver: v2.6.1 (October 29, 2004)

Bonding Mode: fault-tolerance (active-backup)
Primary Slave: eth0
Currently Active Slave: None
MII Status: down
MII Polling Interval (ms): 100
Up Delay (ms): 0
Down Delay (ms): 0

the design was to have 4 physical interfaces bonded to 2 ips for fail-over. actually 6, but the first bond is on different subnet so no issue.. :)

Re: Bonding Problem

Alexander Chuzhoy — Tue, 04 Dec 2007 07:11:36 GMT

What you see is normal.
If you'll bring down and then up 1 interface from the bond0 - just wait a couple of minutes and then try to "restart" the second interface - you won't run into delay.

Your problem is 2 interfaces in the same network. Kernel needs to check/redefine the route to the network you work on.

Re: Bonding Problem

Alvin_14 — Sun, 09 Mar 2008 02:10:47 GMT

Closed