Computer gets IP from DHCP server but has no internet connection

Question

I've installed an isc-dhcp-server on Ubuntu 12.10 and I'm trying to setup a DHCP server on a head node for six worker computers in the local network. The head node itself is a DHCP client receiving the IP-address 192.168.20.1 (on eth0) from an other computer in the network.

IPv4 IP forwarding is enabled on the head node. In /etc/sysctl.conf the following line was added:

net.ipv4.ip_forward = 1

The following rules are set in /etc/rc.local:

/sbin/iptables -P FORWARD ACCEPT
/sbin/iptables --table nat -A POSTROUTING -o eth0 -j MASQUERADE

Furthermore, INTERFACES="eth0" is set in /etc/default/isc-dhcp-server.

The dhcpd.conf file contains (for the moment only one worker computer is in the file):

ddns-update-style none;

default-lease-time 3600;
max-lease-time 7200;

authoritative;

subnet 192.168.20.0 netmask 255.255.255.0 {
  range 192.168.20.2 192.168.20.200
  host hostName {
    hardware ethernet macOfHost;
    fixed-address 192.168.20.20;
  }
  option subnet-mask 255.255.255.0;
  option broadcast-address 192.168.20.255;
  option routers 192.168.20.1;
}

The worker node effectively gets the IP address 192.168.20.20, but it has no internet connection. A ping to 192.168.20.1 is successful, as well as a ping to the computer that serves as a DHCP server for the head node (a ping to 192.168.1.1). The problem doesn't seem to have something to do with DNS as a ping to an IP-address (such as 8.8.4.4) fails.

Update

The network topology is as follows. There is switch connecting a computer (with internet access and running a DHCP server, IP address 192.168.1.1) and 7 other computers. One of these 7 computers gets an IP from the 192.168.1.1 computer. The IP it gets is 192.168.20.1 and the internet on that computer works fine. Now we have a DHCP server running on 192.168.20.1 in order to provide internet access to the other 6 computers, but that fails. They get IP addresses but have no internet access. We are not allowed to modify anything on the 192.168.1.1 node so it should be feasible to make internet work with this setup.

Does someone know what the problem could be?

Output of /sbin/route -n on the client:

Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         192.168.20.1    0.0.0.0         UG    0      0        0 eth1
169.254.0.0     0.0.0.0         255.255.0.0     U     1000   0        0 eth1
192.168.20.0    0.0.0.0         255.255.255.0   U     1      0        0 eth1

Output of sudo iptables -L -v -n: Chain INPUT (policy ACCEPT 2146 packets, 1551K bytes) pkts bytes target prot opt in out source destination

Chain FORWARD (policy ACCEPT 59 packets, 3762 bytes)
 pkts bytes target     prot opt in     out     source               destination         

Chain OUTPUT (policy ACCEPT 1908 packets, 213K bytes)
 pkts bytes target     prot opt in     out     source               destination

Output of ip ro sh default via 192.168.1.1 dev eth0 proto static 169.254.0.0/16 dev eth0 scope link metric 1000 192.168.0.0/16 dev eth0 proto kernel scope link src 192.168.20.1 metric 1

Solution

Everything is working now. It turned out to be a DNS problem; changing the option routers to option routers 192.168.1.1 fixed the issue. Pings must have been blocked somewhere in the network as responses never came through, but surfing in the browser to a specific IP-address was possible, which made me realise that DNS was the problem. Thanks everyone for your help.

Comments

[umläute]

just to be sure: can your router (192.168.20.1) ping the outside world?

[user2611216]

Yes, it can. (10 characters)

[Doon]

what is subnet mask on the 192.168.1.x network? if you are getting an ip of 192.168.20.1 from that network the subnet mask is probably 255.255.0.0? and it looks like you are trying to nat with overlapping subnets. if you change your downstream dhcp config to use a different set of RFC1918 IPS does it work (10.10.0.0/24 for example).

[user2611216]

The subnet mask of that network is indeed 255.255.0.0.

[user2611216]

@umläute: I've posted the output in my question.

[Doon]

also you can't have 2 dhcp servers giving out different information on the same Layer 2 network Just doesn't work and int looking at your config are you trying to NAT on the same interface, it will not work. what I would do is this. move the 6 worker nodes to a different layer 2 network, and hook them into eth1 on the head node. Give them a different RFC 1918 address (10.x ) and then all of your stuff above would work. BUt with only 1 interface and the overlapping subnets/ 2 dhcp servers it will not work.

[Doon]

sorry missed the eth1 above. The problem is your ip scheme (and that you have 192.168.20.1 on both eth0 and eth1). Change your subnet on eth1 and you should work fine.

[user2611216]

IP forwarding was already enabled. I've made this more clear in my question now.

[umläute]

as i said in my answer, just having the line in the config-file will not enable ip-forwarding until the machine is rebooted. what does sysctl net.ipv4.ip_forward output? can the router see the internet?

[user2611216]

The machine has already been rebooted many times since that line was added. The head node (192.168.20.1) can see the internet, the DHCP client of the DHCP server running on that computer cannot.

[ALex_hha]

Could you show output of the iptables -L -v -n and ip ro sh ?

[user2611216]

I've put the output in my question.