Openmptcprouter: [FIXED] Fix for long-lived VPN connections to be fully aggregated at all times

[uryupinsk]

Expected Behavior

Long-lived connections such as VPN site-to-site or SSH sessions must be aggregated above all links at all times.

Current Behavior

With the default OMR and Linux network stack settings, long-lived connections are sporadically not fully aggregated.

Possible Solution

Some posts on MPTCP Github issues and the Red Hat Knowledge Base suggest that the /proc/sys/net/ipv4/tcp_retries2 default value is too high. This setting controls how many retries TCP makes on data segments.

Red Hat suggests that in a High Availability (HA) situation this settings is best decreased to a value of 3.

Steps

1. SSH to OMR
2. echo "net.ipv4.tcp_retries2=3" > /etc/systctl.conf
3. Go to Network > MPTCP and increase Multipath TCP SYN retries to 3
4. Reboot OMR

Context (Environment)

After I changed these settings for a few days, on a Wireguard site-to-site VPN, all links are fully used at all times.

I did not encounter any issues yet with my daily usage.

Specifications

• OpenMPTCProuter version: 0.54.7
• OpenMPTCProuter VPS version: 0.1015
• OpenMPTCProuter platform: x86_64

[Ysurac]

Thanks I will test that and, if there is no issues, this will be in next release.

[hewenhao2008]

@uryupinsk "After I changed these settings for a few days, on a Wireguard site-to-site VPN, all links are fully used at all times"
Are you sure the Wireguard VPN works for all links? I think wireguard works on udp only, and will not bonding all links. How do you make wireguard bonding all links?

[uryupinsk]

@hewenhao2008 UDP aggregation is done by glorytun-tcp by default on OMR. There is also the newer DSVPN available but it is buggy on 0.54.7. I will try it again on the next OMR version.

[uryupinsk]

@Ysurac Have you included this fix in OMR 0.55?

[Ysurac]

Yes it's available in 0.55

[uryupinsk]

Excellent, thanks a lot.