[Dnsmasq-discuss] Bonding interactions with dnsmasq: ip address allocation

[Simon Kelley]

A few weeks back, someone came up with a valid use-case for being able
to allocate IP addreses sequentially, and this looks like a second one.

If you use

http://www.thekelleys.org.uk/dnsmasq/test-releases/dnsmasq-2.58test5.tar.gz

and the new

 --dhcp-sequential-ip

flag, then the machines will get sequential addresses which should, I
think, solve the problem.

HTH

Simon.




Paul Smith wrote:
> Hi all.  I have a system where the server (running dnsmasq) talks to
> clients over a bonded network link using mode 6 bonding.  The clients
> get their IP address information from the server running dnsmasq.
> 
> I don't want to get into the details of mode 6 bonding but suffice it to
> say that the result for us is that which leg of the bond is used ends up
> being based solely on the IP address; if the IP address (as a 32bit
> number) is even then the traffic goes over one leg and if the IP address
> is odd, it goes over the other leg.
> 
> For systems with a lot of clients the default "random" assignment of IP
> addresses works fine; it tends to balance out.
> 
> For systems with fewer clients, life is not so good.  For a system with
> 6 clients if we get an even 3/3 split between even and odd IPs, then
> things are great.  However, if we happen to get a 4/2, 5/1, or even 6/0
> split then the traffic is badly out of balance on the bond and our
> performance takes a severe hit.
> 
> 
> I'm assuming that there's nothing in dnsmasq today that would help me
> with this problem.  Does anyone have a thought about how to address
> this?  One option, of course, is to get the source (today I use the
> default dnsmasq provided by Red Hat) and make some changes myself that
> enforce an even/odd distribution on IP addresses handed out by the
> lease.
> 
> Other options like predefining the IP address based on MAC or whatever
> aren't really feasible (I don't know all the MACs and clients can be
> replaced at any time).
> 
> Ideas/thoughts/alternatives?
> 
> 
> Thanks!
> 
> PS. Yes I know mode 6 bonding is sub-optimal; unfortunately I need to
> PXE boot over this link and other bonding modes don't work with PXE
> unless your PXE driver (provided with the BIOS on the hardware
> remember!) has support for it... *sigh*
> 
> 
> _______________________________________________
> Dnsmasq-discuss mailing list
> Dnsmasq-discuss at lists.thekelleys.org.uk
> http://lists.thekelleys.org.uk/mailman/listinfo/dnsmasq-discuss
>