Re: [SLUG] channel bonding

From: John Pugh (jpugh@novell.com)
Date: Wed Mar 29 2006 - 15:07:54 EST


>>> On Wed, Mar 29, 2006 at 2:17 PM, in message
<1143659838.7515.23.camel@it001.wackyworld.lan>, mike@wackyworld.tv wrote:
> On Wed, 2006- 03- 29 at 10:58 - 0700, John Pugh wrote:
>
>> > they were before. It seems (from googling) that the native kernel
>> > channel bonding creates conflicts with samba/smb and afp services.
>> > Anybody have any input on this? I'm still googling but it's like
>> > looking for a needle in a hay stack.
>> >
>>
>> This helped me a lot.. http://linux-net.osdl.org/index.php/Bonding
>>
>> I'd take a look with ethereal on both channels...it quickly uncovered
>> an inadvertant routing issue I created.
>> I'm not using netatalk, but samba continues to run as good as it can
>> considering it's smb.
>>
>> JP
>
> Great document John! Thanks. It clears up a few mysteries and opens
> some options including the mii tool that I wasn't aware of. I'm using
> SuSE on our servers and it works great. What was the inadvertent
> routing issue? Was it a default route or something being done by the
> network init script? What capture options / filters did you set in
> ethereal? Are you set up round- robin or tlb or alb?
>
>
> Thanks again!
>
First...my apologies for the munged email above...I'm working with SLED 10 beta8 and evolution/groupwise aren't mixing well at the moment.

Inadvertent routing issue... was my stupidity...I started with DHCP on both nics and somehow managed to keep the route for both NICs once I changed them to be static IPs. I didn't realize what that did until I sucked up some packets with ethereal and did some digging when I noticed that I wasn't getting replies consistently across the primary interface. With ethereal...I suck in everything and then filter on one nic at a time and then go thru the packets to see where they were going. For checking SAMBA, you may wish to filter on Netbios session service (I think) and see if you can find any anomalies. Leaving the capture filter wide open may fill the buffer quick if it's small and you have an active network, but then you can simple use filters to find a ton o' stuff once you've captured a bucket full.

I set it up as mode=1 (active-back) for transmit and left it default for receive (mode=0 - round-robin). It was a simple HA config for

Here's more links...that show simple stuff.
http://linux-ha.org/IpFailoverChannelBonding
http://de.lunar-linux.org/lunar/mirrors/README.bonding pushed a pop-up to me and it's just the bonding readme
http://www.linuxhorizon.ro/bonding.html - good stuff on the bonding details

JP

-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.



This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 20:06:42 EDT