5

Setup Description

I have a TP-Link TL-R470T+ Load Balancing router setup at my home. Here is a link to the entire setup along with the routers connected etc. As can be seen from the diagram, I have three different ISPs (for extra redundancy) - all PPPoE connections.

Problem

The main reason I purchased the Load Balancer was that we wanted an uninterrupted connection. However, the connection often goes off for a few seconds to a couple of minutes, kicking users out of zoom call, logging them out of remote servers, etc. Currently, we've set it up in Load Balancing mode. Link backup hasn't been enabled because that causes only one ISP to be active at a given time. However, we also require the bandwidth, as at any point in time, there are approximately 20 devices connected.

Possible Issues

The possible issues that I can think of are

  1. ISP failure - Very less probable as all thee ISPs will need to fail at the same time
  2. Load Balancer Configuration - Some setting in the Load Balancer Configuration I'm missing?
  3. Too many routers/Devices connected - The official documentation of the Load Balancer (Link here) says the device can support up to 10,000 connections, while we have only around 20-25 devices connected
  4. An issue with the router configuration? - All the routers are set up with the following configurations
    • Mode - access point mode
    • SSID - all have the same SSIDs and passwords
    • Channel Interference - Shouldn't be an issue, as the routers are on different floors, and different ones have been configured for different channels.

Another thing I'd like to know is is it impractical to expect a seamless connection from a load-balancer? (E.g. No dropped frames in a video call even when one of the ISPs goes off). If so, is there a better alternative?

I've called the official technical support group multiple times. They suggest some tweaks like updating firmware, changing the DNS, etc, but most of them haven't produced any good results. Any help would be appreciated! Thanks in advance!

PS: I'd posted this Question on the Network Engineering Stack Exchange, but was redirected here!

Edit

Adding some of the suspicious entries of the system log

Module Level Content Comment
Firewall WARNING Detected Ping of Death attack. Dropped x packets. (This occurs quite often) This happens often, but is quite spaced out
PPPoE Client WARNING WAN2: PPPoE sending PADI timeout. Happens once in a while
Firewall NOTICE Detected ARP Conflict. Dropped x packets This occurs rarely, but when it occurs, there are multiple similar logs, in a small time interval, leading to loss of many packets
Firewall Warning Detected suspected ARP attack packets. Be careful Don't know what to make of this. Comes once in a while

Hope this can throw some additional light!

6
  • 2
    I'm a bit curious about your "wireless router" configuration. Through which port are they connected to your central switch? Are they all in bridge mode? Did you end up with each router having a separate subnet? – user1686 May 8 at 17:09
  • All the routers are configured as access points with DHCP disabled. They're all connected through one of the LAN ports (the WAN port is empty). – DDDAD May 11 at 6:12
  • About the subnets, I'd left the subnet masks to their default configurations. Which happen to be 255.255.255.255 for the load balancer and 255.255.255.0 for the routers. – DDDAD May 11 at 6:13
  • @ddad, is this a point to point link? 255.255.255.255 doesn't seem right. If you have different ISPs for the wan links, dropping the TCP states from the lost connection is to be expected. – Tim_Stewart May 11 at 13:56
  • 2
    255.255.255.255 looks perfectly fine for PPPoE (which is indeed a point-to-point link), though I think @DDDAD is mixing up which subnet mask they're looking at...or assuming that there's only one. (If the router has 1 LAN and 3 ISPs, that's 4 networks in total, and 4 subnet masks!) – user1686 May 11 at 16:37
8
+25

I'll only answer this part first:

is it impractical to expect a seamless connection from a load-balancer? (E.g. No dropped frames in a video call even when one of the ISPs goes off). If so, is there a better alternative?

With a basic "load balancer" this is very difficult to achieve as far as I know, simply because each of the 3 ISPs requires you to use a different IP address to access the Internet.1

So whenever the link through ISP 1 goes down, all TCP connections made using that ISP's IP address are as good as gone – there is no way to migrate them to a different endpoint at all, much less without losing any packets. The same also applies to most UDP-based applications (there are a few which support endpoint roaming, but most video streaming protocols do not).

It's quite possible (though I have no way to confirm) that it is actually an unexpected ISP switch that causes all existing connections to die. You should watch your router's system logs or check what is being reported as your "public" IP address from the affected computer.

One possible workaround is to set up a VPN that uses a roaming-capable protocol (WireGuard in particular works well), and always make the video calls through that VPN. This works because your video connection will be using the VPN server's IP address the whole time, regardless of which ISP you're using to reach that server.

(But even with a roaming-capable protocol, there can still be some packet loss – if the link to the current ISP dies (i.e. switchover is unexpected), the server will not be aware of what just happened on your end and will just continue sending packets to your old address until it gets informed of the new endpoint. The protocol needs to be designed to buffer and retransmit if necessary.)


1 (Large enterprise networks use BGP to share the same IP addresses across multiple ISP connections, but this can only be done with the ISPs' cooperation and has quite a few additional costs.)

2
  • Interesting bit about the vpn roaming. I have never heard this before. – Tim_Stewart May 11 at 17:43
  • Thank you very much for the VPN insight. I'm yet to experiment with that though. I'm not sure if this makes sense, but could it possibly be because of traffic overload? – DDDAD May 13 at 16:00
3

Based on your diagram, I'm assuming that the LOAD BALANCER is the TP-Link TL-R470 router

Each ISP1 will typically assign you an IP, it can be dynamic or fixed, in any case you need to setup route(s) from those IPs to your inner network infraesctructure, unless this is done automatically by the router.

enter image description here

Usually your inner network has what is called private IP address and the ISP could be configuring your connection as public IP address, you can check that with your provider, anyhow that is the router function to connect network with different configurations

According to the router manual you can follow the

  • section 4.5.4 -> load balancing
  • section 4.5.5 -> routing

Here you can download the router manual

enter image description here

Also a few people have posted videos on youtube just look for "TP-Link TL-R470 load balance" and you can get more details about this.

If this is not clear or lacks more information let me know

6
  • 1
    Thank you for the answer @Mauricio, however, I've already done all the things mentioned in the manual. I already have the load balancing configuration setup. The problem however is that it goes off for a minute or two at times, and I'm not able to diagnose the source of the issue! – DDDAD May 13 at 6:59
  • check if your router has logging capabilities and enable them to gather more information. – Mauricio Gracia Gutierrez May 13 at 14:09
  • 1
    I've added some insights from the system log report as an edit, hope that helps – DDDAD May 13 at 16:14
  • ARP = Address Resolution Protocol, the two ARPs errors that you included could be related, this is probably because a segment of your network has an IP range that conflicts with your existing configuration – Mauricio Gracia Gutierrez May 13 at 19:56
  • Regarding the Ping Of Death, check the firmware of the router has a new version - community.tp-link.com/en/business/forum/topic/169736 – Mauricio Gracia Gutierrez May 13 at 19:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.