I have two servers at Linode running OpenVPN. One is in New Jersey and the other in California. The NJ server is running CentOS 5.6 while the CA server runs CentOS 6.3. However I have the same Repoforge OpenVPN RPM, openvpn-2.2.2-1, installed on both. Both have comp-lzo enabled, though the CA server has version 2.03-3.1 and the NJ server has 2.06-1. The connection works fine, but I still get the code=22 error in syslog. The 5.6 server in NJ maintains tunnels with half a dozen remote hosts. Only this connection generates the error. Another remote runs Debian 5.0.8 with OpenVPN 2.1~rc11-1, yet it does not cause the code 22 error. It is also running version 2.03 of lzo.
Every solution I've read points to comp-lzo as the culprit, but that cannot be the case here unless the difference between versions 2.03 and 2.06 of lzo is the problem. Still if that mismatch mattered, I'd see the same errors from the Debian connection. As I say, the connection between the Linode servers works fine, but I still get these errors. Any hints? An MTU issue perhaps?
I've run OpenVPN for quite a few years now without incident. I'd like to fix whatever causes this problem so it doesn't fill my syslog. These are simple static-key tunnels; the configuration on the NJ server reads like this:
Code: Select all
dev tun
ifconfig 10.1.1.1 10.1.1.12
secret /etc/openvpn/keys/my.key
port 51002
user nobody
group nobody
comp-lzo
ping 15
ping-restart 45
ping-timer-rem
persist-tun
persist-key
verb 3
The configuration in California has a "remote" directive and, of course, the IP addresses in the "ifconfig" directive are reversed.
