OpenVPN Adapter Address ICMP Behavior

Sometimes OpenVPN will not respond to ping on certain virtual addresses used solely for routing endpoints when using the net30 topology. Do not rely on pinging the OpenVPN endpoint addresses as a means of determining if the tunnel is passing traffic properly. Instead, ping something in the remote subnet, such as the LAN IP of the server.

According to the OpenVPN FAQ, in the section titled Why does OpenVPN’s “ifconfig-pool” option use a /30 subnet (4 private IP addresses per client) when used in TUN mode?:

As 192.168.1.5 is only a virtual IP address inside the OpenVPN server, used as an endpoint for routes, OpenVPN doesn’t bother to answer pings on this address, while the 192.168.1.1 is a real IP address in the servers O/S, so it will reply to pings.

This may seem a little counter-intuitive, since on the server something like this is seen in the ifconfig output:

tun0: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> metric 0 mtu 1500
        inet6 fe80::202:b3ff:fe03:8028%tun0 prefixlen 64 scopeid 0xc
        inet 192.168.100.1 --> 192.168.100.2 netmask 0xffffffff
        Opened by PID 27841

While the client shows:

tun0: flags=8051<UP,POINTOPOINT,RUNNING,MULTICAST> metric 0 mtu 1500
        inet6 fe80::202:b3ff:fe24:978c%tun0 prefixlen 64 scopeid 0xa
        inet 192.168.100.6 --> 192.168.100.5 netmask 0xffffffff
        Opened by PID 1949

In this case, .5 or .1 will not likely respond to ping. .5 because it’s a virtual address, and .1 because the client has no route to reach it directly. The .5 and .6 addresses are part of a /30 that goes from .4 to .7, and trying to ping .1 would go out the default route instead.

There are many cases where the far side of an OpenVPN tunnel can respond to ping, but not the local. This is also counter-intuitive, but works especially in cases where there is a site-to-site link. If the server shows its tun addresses as “x.x.x.1 -> x.x.x.2” and the client shows the reverse - “x.x.x.2 -> x.x.x.1”, then the far side will likely respond to ping from both ends.

In contrast, when using “topology subnet” these virtual addresses and /30 subnets are not used so these issues are not present.