OK I think I may have figured out what is going on.
I believe the NetBalancer module may be trying to split my traffic out over multiple links, which is why with UDP (connectionless) it is dropping the packet, however with TCP it is re-trying and is getting the packet through, but causing high latency because of the delay.
I definitely have configured the VPN tunnels to use a specific interface through netbalancer.
I was hoping someone, or fulvio, may be able to assist with how Net Balancer should be configured in certain situations.
For example, a scenario where I have a zeroshell at both sites, and two ADSL links of identical speed at both ends, which only purpose will be for a VPN bond to load balance traffic across, how would I have Net Balancer configured? Load Balance + Failover, with both gateways having an equal weight of 1?
Or for example in the configuration where I have a head office with two links, want all outbound traffic to go via one link, and then have a remote branch with a VPN Bond to both links for load balance, would I have the NetBalancer configuration set to Fail Over, with the primary link weighted as 2, with the secondary link weighted as 1, and then the VPN tunnels configured to go out via each link?
This is the configuration I have been trying, with mixed results as mentioned above in regards to packet loss and high latency.
I have tried Beta 11 & Beta 12, thinking that I may have uncovered a bug, but I get the same result in both, and can re-produce my results at different sites, with different links.
In the first example I gave, I have had to remove the bond and go back to a single VPN tunnel using ‘auto’ as the default gateway and only using one link, and when configured as a UDP OpenVPN tunnel, the latency problem does not exist. It only occurs when I move into the Bond / Net Balancer combo does it return.
Hoping someone can shed some light, as I really need to get these networks solved and my bonds completed!