Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 24 Jan 2001 09:36:18 -0700 (MST)
From:      Nick Rogness <nick@rapidnet.com>
To:        Andreas Brodmann <andreas.brodmann@gmaare.migros.net>
Cc:        Dejvid Zaninovic <dzaninov@mediaondemand.com>, freebsd-hackers@freebsd.org
Subject:   Re: IP Address Overtaking
Message-ID:  <Pine.BSF.4.21.0101240926080.68311-100000@rapidnet.com>
In-Reply-To: <3A6EF007.9F06DBF8@gmaare.migros.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 24 Jan 2001, Andreas Brodmann wrote:

> 
> On normal internetworking hosts, without the necessity of high availability
> this works fine. Not all hosts do update or even flush their arp cache with
> the same frequency though. Some have a cycle of less than one minute on
> routers on the other hand the default arp cache timeout is a lot higher which
> would force clients not in the same subnet to wait until the router flushes
> its arp cache until they can access your FreeBSD machine again.
> -> not ha compliant.

	The time it takes to flush is very small.  During that time the
	router queue's up the request and waits for a reply.  Once the
	router has it, everything is transparent.

	I would not recommend playing with MAC addresses at all.  Switch
	things using IP and let the ARP protocol take care of itself.

> There is a way to solve this problem by having a second interface in each
> cluster
> partner serving as standby interface. To this interface you assign the mac of
> its
> partner's interface and all its interfaces ip addresses.
> 
> Just a hint: Have a look at scyld.com and Donald Becker's new Linux driver
> architecture. Many new cards allow for using more than one mac per card
> even without going into promiscuous mode. They can then be assigned to
> different subinterfaces. I don't know wheter the FreeBSD drivers support
> this. Anyway we still keep to the old fashioned way mentionned above, as the
> new Linux network driver architecture is not yet as stable as it could be, but
> once it is this would solve your problem.

	I think this is a bad idea in a clustering enviroment.  You are
	taking the job of a switch and moving it to the card/software by
	fiddling with MAC addresses on the hosts.

	I guess I can see where this may be useful (trunking) but taking
	over the MAC could cause problems...like duplicate MAC's etc,etc.

	Of course, this is my opinion and I could be wrong.

Nick Rogness
- Keep on routing in a Free World...  
  "FreeBSD: The Power to Serve "



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.21.0101240926080.68311-100000>