Hi Liam,

I checked the code again and confirm that preempt doesn't pass the message
of one carp going down to other carps installed on the same physical
interface. Thank you for your effort in duplicating the issue.

The bugs are about ARP balance. I haven't claimed to OpenBSD or others
because the project is very tight now. I'll do it shortly.

If carp0 on fw1 spi4 is down but carp1 on fw1 spi7 is still master, the
traffic is also down. What I expect is that when carp0 on fw1 spi4 is down,
carp1 on fw1 spi7 will flip to BACKUP and carp 0 and 1 on fw2 go to MASTER.
If it happens, the traffic could be continued. Otherwise, the traffic stops.

Regards!

George Chen 

-----Original Message-----
From: tech-kern-owner@NetBSD.org [mailto:tech-kern-owner@NetBSD.org] On
Behalf Of Liam J. Foy
Sent: 2006 05 11 16:35
To: George Chen
Cc: tech-kern@netbsd.org
Subject: Re: question about CARP

On 11 May 2006, at 05:07, George Chen wrote:
> Sorry, there are so many lines.:)
> When I bring down carp0 on fw2 by 'ifconfig carp0 down', carp0 on fw1
> becomes MASTER, but carp1 on both sides remains unchanged.
>
> Sorry I didn't sync to head version, didn't use your patch neither.  
> Instead
> I checked your patch. My opinion is that this is a design bug. The  
> reason is
> that there is no mechanism to transfer the message of one carp  
> going from
> backup to master to all other carps on the same firewall. Without the
> message, how can other carp react to it?

No, preempt should do this. I'm going to try and get my network setup so
I can try. I'll let you know.

>
> What is my modification? There are bugs in arp balance, which also  
> exists in
> your patch. Additionally I'm making carp support for layer 2 HA and  
> bridge
> balance.

What are the bugs in arp balance? Have you let OpenBSD know of these  
bugs
you claim to have found?

> Cheers!
>
> George Chen

Cheers,
		---
		Liam J. Foy
		<liamjfoy@netbsd.org>
		<liamfoy@sepulcrum.org>
		BSDPortal.org