Huge lag spikes (5s+) every 30 seconds when connected to Tailscale

[#1]

I am having a strange issue that I have narrowed down to a combination of using Tailscale and Plusnet. Every 10-20 seconds I get lag spikes upward of 4-5 seconds and huge packet loss (ping and MTR attached).

• This only happens when connected to my work's Tailscale network
• No one else using our Tailscale network has this issue
• I have had this issue on a Windows and a Linux machine
• The issue goes away when I disconnect from Tailscale
• The issue only occurs at home on a Plusnet connection / router; using my mobile broadband or taking my machine to work doesn't show the same issue

Is there anything I can do to resolve this? I am going to have to move provider if not, because it is unusable when working from home

[#2]

One thing that may be worth checking is the MTU - the maximum payload size will be lower when using a VPN due to the extra headers, and if something along the path doesn't handle the size it will split the packet into two and not every recipient will reassemble split packets. That can lead to packet loss and strange delays. 

Lowering the MTU by 10 or 20 just may help.

[#3]

Thank you for your suggestion, for now I have requested a static IP which seems to have resolved the issue as I am no longer being bounced through 133.hiper04.sheff.dial.plus.net.uk which is where all the packet loss was occurring.

Annoying, but worth the £5!

[#4]

Can I ask you how you determined that the packet loss was occuring at 133.hiper04.sheff.dial.plus.net.uk?

[#5]

If you look at the MTR screenshot el the third line says "waiting for reply"; occasionally it would get a reply which resolved to that 133 relay, and it was experiencing 90%+ packet loss, which correlated with the lag spikes.

[#6]

As far as I know 133.hiper04.sheff.dial.plus.net.uk rarely responds to ICMP/ping traffic, therefore not inidcative of a problem.

Whilst I can't explain your observations following the application of the static IP, your MTR shows poor ping response times from the first hop, which I assume to be your router? Where this is the case, I'm normally inclined to say that the problem is local to your network.

[#7]

I'm not really sure what to tell you, I am running exactly the same equipment / setup as I was yesterday, connected to the Tailscale VPN, and I have just run an MTR with 600 packets again and had 0.0% loss with a highest average ping of 10.7.

The only difference is today I have a static IP, and yesterday I could reliably cause the issue to occur / resolve by switching the VPN on and off. I have also been unable to replicate this issue on any other ISP. There appears to be some inherent incompatibility between how certain software (including Tailscale) works, and how Plusnet's relay servers work, that is avoided by bypassing the hop with a static IP.

Others appear to have similar issues / results with different software, it's just frustrating the Plusnet refuse to acknowledge this as a problem; I contacted support and they basically said "it's a VPN software issue" and hung up on me.

https://community.plus.net/t5/Fibre-Broadband/New-line-packet-loss-at-Plusnet-Hop/td-p/1748064

https://www.google.com/search?q=133.hiper04.sheff.dial.plus.net.uk

[#8]

---

@vidoardes wrote:

... it's just frustrating the Plusnet refuse to acknowledge this as a problem; I contacted support and they basically said "it's a VPN software issue" and hung up on me.

https://community.plus.net/t5/Fibre-Broadband/New-line-packet-loss-at-Plusnet-Hop/td-p/1748064

https://www.google.com/search?q=133.hiper04.sheff.dial.plus.net.uk

---

Ignoring the unacceptable call termination, those links are probaly not far off your classic case of correlation versus causation.

In your traces there was packet loss to your gateway. Unless I'm misinterpreting things, there's no way anything out on the Plusnet network can influence that.