Re: Linux NIC bonding

  • From: Dan Norris <dannorris@xxxxxxxxxxxxx>
  • To: exriscer@xxxxxxxxx
  • Date: Wed, 12 Dec 2007 05:57:02 -0800 (PST)

As I mentioned, I use active-passive (aka active-backup). Sorry if my 
terminology confused you.

Dan

----- Original Message ----
From: LS Cheng <exriscer@xxxxxxxxx>
To: dannorris@xxxxxxxxxxxxx
Cc: Oracle L <oracle-l@xxxxxxxxxxxxx>
Sent: Wednesday, December 12, 2007 4:33:44 AM
Subject: Re: Linux NIC bonding


Hi Dan

What mode do you use, 0, 1, 2.. etc? Sounds you have used 1 (active-backup). 
This is what kernel doc says about it:

active-backup or 1


        Active-backup policy: Only one slave in the bond is

        active.  A different slave becomes active if, and only
        if, the active slave fails.  The bond's MAC address is

        externally visible on only one port (network adapter)

        to avoid confusing the switch.


        In bonding version 2.6.2 or later, when a failover
        occurs in active-backup mode, bonding will issue one

        or more gratuitous ARPs on the newly active slave.

        One gratutious ARP is issued for the bonding master
        interface and each VLAN interfaces configured above

        it, provided that the interface has at least one IP

        address configured.  Gratuitous ARPs issued for VLAN
        interfaces are tagged with the appropriate VLAN id.


        This mode provides fault tolerance.  The primary

        option, documented below, affects the behavior of this
        mode.


I have always used 0 in the past

Thanks

--
LSC


On Dec 12, 2007 6:22 AM, Dan Norris <dannorris@xxxxxxxxxxxxx
> wrote:
I'm building a cluster using Oracle Clusterware and have configured redundant 
pairs of NICs on the Linux servers using the Linux bonding module 
(active-passive). I know how to configure the Linux end of things and that's 
working fine. However, the switches used by this customer (they're Dell 
branded--not sure who really makes the guts) have been getting "confused" by 
bonding and our simple ping tests lose about 50% of the packets when we plug 
the redundant pairs into the two separate (trunked) switches. 


This is only peripherally related to Oracle--it's really a networking and 
server config issue, but I'm hoping that someone with more networking 
background than I have can explain what switch configuration will enable this 
to work properly. The customer has stated that this will never work (though I
 have done it before at other sites and it worked fine). The customer has 
stated that the Linux configuration must be two active NICs
 with a virtual IP and virtual MAC. I imagine that will work, but I'm not 
familiar with that configuration and I don't think Linux has native support for 
it, so it'd probably be a Broadcom or Intel software package. It's questionable 
how Oracle would view that from a support perspective (if they care at all). 


So, the question: Is there some "magical" switch setting that would possibly be 
missing in a fairly "default" configuration that might enable this to work 
properly? I figure it's a long shot asking a networking question on the Oracle 
list, but thanks for any pointers.


Dan








Other related posts: