LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: hard wired solution to arp problem

To: Julian Anastasov <uli@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: hard wired solution to arp problem
Cc: lvs-users@xxxxxxxxxxxxxxxxxxxxxx
From: Joseph Mack <mack@xxxxxxxxxxx>
Date: Sat, 27 Nov 1999 01:25:39 -0500 (EST)
On Sat, 27 Nov 1999, Julian Anastasov wrote:

Hi Julian, are you up real early or real late. I normally get up at 4:30am
and so am in bed by about 9:30pm and so normally would not be up now, but
am shifting my circadian rythm back an hour a day to match Australian
time.

> > 
> > I have assumed that the realservers do not ask "who has ROUTER tell VIP"
> > - I thought they only asked "who has ROUTER tell realserverIP".
> 
>       This is incorrect assumption. If You apply one of the two patches

one of your two patches (not Stephen's and yours)?

> You will see in the kernel log that real servers send ARP queries with
> SIP=VIP (who-has LVS tell VIP).

you have the switches on in the patch?

I haven't looked real hard for these arp requests with tcpdump
as the VIP was on wierd devices that I didn't think tcpdump
listened too (eg tunl0 or lo:0). I will go look.

> > The realserver has a packet src=VIP,dest=ClientIP but
> > when it asks for a route to ClientIP it does so from
> > the realserverIP (I thought). The LVS works fine for
> > me if there is no route for the VIP on the realservers.
> 
>       No. It works because there are no other hosts to ask "who-has VIP
> tell HOST". 

Why does a realserver what to know who has the VIP? Doesn't 
each realserver think it has the VIP and not need to ask 
(I don't see any arp table entries for the VIP on the
realservers).

> But the real servers always ask "who-has LVS tell VIP" and may
> be reply is not send from LVS as the VIP is configured in LVS too.

I take it that what you are calling the LVS, I am calling the director (I
call the LVS=director+realservers). In this case are you saying that the
realservers are asking for MAC address of the director? Why are
they doing that? 


> 
> > > In this configuration it is again not possible WEB1
> > > for example to switch to LVS mode when LVS box is down.
> > 
> > I'm sorry I don't understand this sentence.
> 
>       Sorry, I'm always talking about the more complex configuration
> where any number of the real servers can work as LVS when the host LVS is
> down. But the above configuration works, i.e. one LVS box and no other
> hosts on the LAN which can ask "who-has VIP tell HOST".
> 

Are you saying that there is one director box and no other realservers
on the LVS local LAN which can ask "who-has VIP tell realserver"?


> > do you mean that if a director fails, that a realserver
> > takes over the role of director? That's a neat idea.
> 
>       Yes, it is already working idea. But the ARP must be patched.

this is really neat.

What's going to happen to all of this with the 2.4.x kernels? Are we going
to have to start all over again like we did with the 2.2.x kernels? Is
netfilter (which I know almost nothing about) going to change the arp
problem?

Joe

--
Joseph Mack mack@xxxxxxxxxxx


----------------------------------------------------------------------
LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
To unsubscribe, e-mail: lvs-users-unsubscribe@xxxxxxxxxxxxxxxxxxxxxx
For additional commands, e-mail: lvs-users-help@xxxxxxxxxxxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>