[ILUG] Heartbeat and intelligent switches

[Niall O Broin]

On 6 Jun 2005, at 14:36, Colm MacCarthaigh wrote:

> On Mon, Jun 06, 2005 at 01:49:35PM +0100, Niall O Broin wrote:
>> I realise that without more details about the networking
>> infrastructure, which I don't currently have, it might be hard to say,
>> but I'll ask anyway. In general, can this cause a problem, and if so,
>> how can it be obviated?
>
> The most reliable way is to migrate the mac-address aswell, as part of
> the resilience. We do this for all of our servers, and we have 200ms
> failover with no problems. Though in our case we're using the linux
> bonding driver.

That thought had occured to me, but as far as I understand, the problem 
this organisation has is with the packets from the same IP addresses 
coming in on different ports. However, I could well be misinformed. 
This is for a large European organisation, and on occasion, getting 
accurate technical information through the layers can be a tad tricky.

> If this isn't an option, then the very least the failover should do is
> to trigger a few outbound packets to the gateway after failover occurs,
> to give the network a chance to learn the new MAC address.

That's a plan - I must do that.

>  After that, if the switches/gateway don't play nice, it's arp-timout 
> tuning time :/

Which amounts to saying - "We have an application level issue, and to 
solve it, you're going to have to change the network configuration". If 
I felt that bloody lucky, I'd buy lottery tickets, and to hell with the 
maths.

> That said, it mostly just-works these days.

I hope so. It looks like some testing will be in order anyway - 
trivially easy to test of course, but we might be able to get a few 
quid out of them for it anyway :-)

One part of the organisation wants to have a very important system made 
highly available, so it's a question of persuading the other part to do 
what's needed.



Niall