On Thu, 10 Jul 2008, "Heinz A. Krebs" <heinz@xxxxxxxxx> wrote:
> i'm trying to setup a load balancing system with 2 servers and with
> qemu-kvm-engines running inside. tried it already some days (various
> techniques), but could not succeed, so i'm asking for help, ...
> on Server 1:
> eth0 is bidged to br0, eth1 not used by now
> br0 192.168.0.110
> br0:1 192.168.5.110 (DIP)
> br0:2 192.168.0.40 (VIP)
Just to check - Server1 is the director, yes? Is this a VM, or the
> the server is running qemu-kvm, with a virtual network-card connected to
> br0. the networkcard gets the address 192.168.5.115, with default
> gateway 192.168.5.110. inside the virtual engine i'm running a
Again, to be clear - the "realserver" (in LVS terms) is a QEMU VM?
> relevant part of my ldirector.cf:
> real=192.168.5.115:53 masq
> and a similar entrie for ssh
You'd be much better off just doing one service at a time. Since SSH
responds with a banner (DNS does not) please consider getting your LVS
up and running using SSH alone.
> from the server i can:
> ssh 192.168.5.115 -> connection o.k.
> telnet 192.168.5.115 53 -> connection o.k.
Which means that the bridge device is working OK, at least locally.
> but from outside (CIP 192.168.0.5):
> ssh 192.168.0.40 -> no route to host
> telnet 192.168.0.40 53 -> connection refused
> ping 192.168.0.40: ping is o.k.
Is 192.168.0.5 a completely separate physical machine?
> tcpdump 'tcp port 53' says:
> 12:37:19.757766 IP 192.168.0.5.34186 > 192.168.0.40.domain: S
> 3157355270:3157355270(0) win 5840 <mss 1460,sackOK,timestamp 150219440
> 0,nop,wscale 6>
> so it seems that the packages reaches the server, but is not forwarded
> to the virtual machine ..
Which machine - real or VM - was that taken from?
Strip it back to SSH only, and try again. This time run a tcpdump on
the director and the realserver (the QEMU VM) and see what you get.
I'm fairly sure there's something basic going on here.