Starlink Bypass Mode not working and 1/4 of expected speed

[Ceroy132]

Expected Behavior

Speeds should be aggregated to near 200 mbps download and starlink should be possible in bypass mode but I cant seem to get it to work.

Current Behavior

All the configuration is the default non-modified from install and instructions from: Install Instructions

I am trying to aggregate DSL + Starlink:

For the DSL I have easily turned off DHCP and set the ipv4 from the instructions to 192.168.10.2 and the gateway to 192.168.10.1 and then done the same on the settings wizard for openmptcprouter. This seems to work fine and in the status page I can see it showing connected with some warnings "Some connectivity tests failed" and "No Server ping response after 1 second" but the internet is working and can't seem to get around these warnings.

For the Starlink, is where the problems are happening. I have tried both router and bypass mode.

In router mode there is no option to disable DHCP or change the gateway / IP. So instead I changed the gateway to 192.168.1.1 and ipv4 to 192.168.1.2 which seems to work although sometimes the device on the status page shows "ip not leased by this router".

In bypass mode, no matter what configuration I tried it would never connect. With ethernet from a pc to the starlink router I was able to have an internet connection and when doing ipconfig in cmd I could see a different gateway at something like 100.64.0.1 which I also tried but to no avail.

So now using Starlink in router mode and DSL then this leads me to a final issue regarding the speeds. If I am using the DSL + Starlink which are each 100 mbps download then I should theoretically receive near 200 mbps download when aggregated but for some reason I am receiving speeds of 50 mbps download using speedtest.net. Is there a different speed test I should be using? How can I debug whats wrong here?

Specifications

• OpenMPTCProuter version: v0.63-6.12 r0+30806-070d8eb4d5
• OpenMPTCProuter VPS version: 0.1052 6.12.73+deb13-arm64
• OpenMPTCProuter VPS provider: Oracle Cloud
• OpenMPTCProuter platform: x86_64 (n97 mini pc)

[Ysurac]

You can use omr-speed-test via SSH to test aggregated speed and omr-speed-test eth1 to test eth1 speed for example

[Ceroy132]

To add to this, I have done a omr-test-speed and checked the MPTCP bandwidth tab:

the omr-test-speed shows a current speed of 8000k
the MPTCP bandwidth tab shows ~10 MB/s from DSL and ~0.06 kB/s from starlink (roughly)

which is kind of weird since prior to doing a reboot on openmptcprouter, it was the reverse where starlink was actually ~10 MB/s and DSL at low kbps values

[Ceroy132]

You can use omr-speed-test via SSH to test aggregated speed and omr-speed-test eth1 to test eth1 speed for example

just did this using wan1/wan2 instead and then in the MPTCP bandwidth tab I can see it using the appropriate wan1 (dsl) or wan2 (starlink) depending on what I used in the command and the speeds seem correct at roughly ~10 MB/s

However when doing just omr-test-speed the results in the MPTCP bandwidth tab show only using the wan1 (dsl) not aggregated so why might that be?

[Ysurac]

I would need a screenshot of status page. It seems you don't have aggregation.

[Ceroy132]

note that the "your ip was not leased by this router" wasn't there in the omr-test-speed I just did above - it only appeared after just rebooting the device again. Everything else was showing the same.

[Ysurac]

The "IP not was not leased by this router" mean current client was not given by OpenMPTCProuter, as you rebooted client didn't asked for IP again. Not a problem.
But the "No server ping response after 1 second" is a problem.

Via SSH on the router, what is the result of ping -B -I wan1 130.61... (VPS IP) ? And what is the latency ?

[Ceroy132]

Regarding the command, there is no response and I have to force cancel it. 100% packet loss with 51 packets transmitted, 0 received, time 51971ms.

So its definitely not able to communicate to the server but why?

[Ysurac]

ICMP is allowed on the server ? Can you try same command to IP 1.1.1.1 ? What is the result of ip r ?

[Network-Traditions]

We have been using Starlink in bypass mode for many years. Currently, the best method for a Starlink WAN is to use a DHCP configured physical interface of the OMR router to connect directly or through an Ethernet switch to the Starlink Ethernet interface. Upon setting Starlink to bypass mode it will provide a Starlink network DHCP ip address to the connected OMR router interface:

Interfaces Tab: Protocol: DHCP client
Carrier: Present
Uptime: 3h 53m 22s
MAC: 5A:86:22:78:7D:39
RX: 6.11 GB (7498388 Pkts.)
TX: 620.27 MB (3983571 Pkts.)
IPv4: 100.104.###.###/10
Devices Tab:

Starlink will renew the DHCP ip address every ~600 seconds; for best results the Ethernet configuration and circuit benefits from low latency and limited managed switch services. In our case, OMR is a virtual machine on an XCP-ng hypervisor connected to a dual 10G LACP connection to a level 2 managed Ethernet switch, which uses a dual 10G LACP connection to a second switch, which makes the final connection to Starlink over a 2.5G Ethernet port. We had issues with this configuration at first because our managed switch was configured with some security services on the LACP ports that caused XCP-ng and the first LACP connection to drop while the extra security process was processed. Upon disabling that feature, the connection became rock-solid.

[Ceroy132]

Thank you for the response regarding the Starlink issue in bypass mode. I had previously tried with it setup in DHCP mode but to no avail. It was a direct ethernet from the starlink router to the OMR router - no switches or anything like that. In the status page I was able to see it grabbing what I assume would be the correct ipv4 and gateway.

I will have a try again though and see if I can figure it out. I don't remember exactly what it would show in the status page but I'll be able to report that here too.

However I need to figure out the problem with the "DNS Issue: can't resolve hostname" first as thats prevented me from doing anything further. I wonder if its related to having updated the openmptcprouter from the sysupgrade tab. Since then I have tried completely re-installing OMR. I might try either a snapshot version or downgrading it to see if it makes a difference.

[Network-Traditions]

I recall making our first connection directly from Starlink Ethernet to our OMR physical X86_64 device and don't believe we used a cross-over cable, but that would be something to evaluate whether the interfaces on each device can accommodate a direct cable connection without a cross-over cable. Also checkout the following thread:

We've been using StarLink V2 in "bypassed" bridged mode for several months now. 2 DHCP servers on the same subnet is undesirable so make sure Starlink is the only DHCP server connected to a subnet where your OMR router is the only WAN DHCP client as the Starlink router will only have one DHCP address available (only one device will succeed in obtaining the DHCP address). Take a look at #2584 with a specific focus on the "Force link" setting (Network->Interface->Edit->Advanced Settings) and also take a look at https://nelsonslog.wordpress.com/2021/04/07/openwrt-vs-starlink-dhcp-leases.

[Ceroy132]

ICMP is allowed on the server ? Can you try same command to IP 1.1.1.1 ? What is the result of ip r ?

okay so I managed to solve the problem, was indeed related to the ICMP being blocked on the VPS. And the other issue I was having was due to the physical interface eth1 and eth2 being swapped around after the reboot.

As for the starlink test in bypass mode, I have not yet gotten to that but will have a look when I get the time tomorrow and report back here. Thank you Network-Traditions for the further follow up with that issue - i'll take a look and try a few different things you've mentioned.