Is it just me or has anyone else noticed that tailscale connection speeds have dropped off a cliff for seemingly no reason in the past week or so? Used to get perfectly acceptable speeds even over cellular with 100-200mbps speed tests (and gigabit symmetrical at home) but now when I’m trying to stream I’m getting like 5-10mbps on a verified direct connection. I also had a weird problem on my computer (windows) where when I have the tailscale client open and try to access another tailscale device (like my NAS drives) it routes lan traffic through it resulting in similar painfully slow speeds, even when the access address is the local IP and not the tailscale IP or magic DNS address, and as soon as I close the client I get full gigabit speeds between devices again. I’ve made precisely zero changes to any options, configs, installs, network/hardware setups, etc for months now, only changes would be version upgrades so I wonder if that might be the culprit.
Edit: it occurred to me to check the log and it’s full of contantly recurring bind errors which is obviously unusual. Any ideas?
002026/05/20 17:06:38 localListener failed to listen on 100.100.200.200:443, backing off: listen tcp4 100.100.200.200:443: bind: permission denied
2026-05-20 17:06:47.097727+00:002026/05/20 17:06:47 localListener failed to listen on 100.100.200.200:80, backing off: listen tcp4 100.100.200.200:80: bind: permission denied
2026-05-20 17:06:51.592011+00:002026/05/20 17:06:51 localListener failed to listen on 100.100.200.200:53, backing off: listen tcp4 100.100.200.200:53: bind: permission denied
100.100.200.200 is the tailscale IP for my truenas box, port 443 and 80 are obvious and 53 is DNS because the server also hosts adguard.
I have had route issues come and go. This morning I finally deleted Tailscale from my desktop computer where it was giving me grief. I use a separate OrangePi as my Tailscale device. Small, cheap, dependable so far, and auto-updating. Don’t ask me how I did it, I need to start from scratch and document it so I have it written down.
A wise man once said “have you tried turning it on and off again". Well, I did, a couple times, but it didn’t do anything. I tried it again just now (just restarting the truenas app) and presto chango now it wants to work (at least the remote access speed) just fine. So idk, but that seems to have fixed it.
This very likely has everything to do with your network configuration and nothing to do with TrueNAS.
Your favourite AI can help you troubleshoot this but specific things to check:
Is the container using host networking? It will work better if yes
Are you being relayed? It will work MUCH better if no.
Is your NAT set correctly? You need endpoint-independent NAT. You can see in the Tailscale web UI – select your file server node, then scroll down to the bottom of the page and look under CLIENT CONNECTIVITY. It should say: Varies: No.
Do you have IPv6 available from your ISP (DHCPv6-PD) and enabled on TrueNAS? If will work better if yes. Under CLIENT CONNECTIVITY it should say IPv6: Yes if you have available and working.
You do NOT need to open incoming ports / port forwarding on your firewall if you do the above.
Unless you explicitly set a port, the Linux Tailscale daemon uses a random port so it won’t match your port forwarding. You can confirm using netstat, and can set a fixed port via environment variable:
- TS_TAILSCALED_EXTRA_ARGS=--port=41641
It is very hard to remotely troubleshoot performance problems in a system that only you can see. Understanding and respecting that you don’t use AI but after going over the most common issues, it will likely be up to you to figure out why the problem is happening and then someone here may be able to help address.
The desktop/mobile clients use 41641 but the Linux client defaults to a random port.
Port forwarding can be helpful if the NAT is not endpoint independent; otherwise with endpoint independent devices will very quickly connect to each other directly.
If the port number was the issue, it simply wouldn’t work at all.
Given the little info and the fact that it is “random” (which it isn’t!), this looks like networking/routing or being relayed (or using tailscale node as “exit node”?).
There were “no changes”, which only make it more confusing.
Correct, you don’t need open ports provided you have a proper (endpoint independent) NAT.
I don’t know why @daktah is having problems and he/she stopped responding. Based on the triage provided, Tailscale is not relaying, so it should work well. For comparison, I get >800Mbps over tailscale between the TrueNAS server (14 year old low-end server) and a Pi 5, essentially wire speed on gigabit Ethernet.
i havent replied because it has not improved and no one has anymore advice so far. its still doing the exact same thing with the exact same problem, no i dont have any exit nodes or any other advanced settings enabled on any device in my tailnet either. ive pretty much given up as it usually works well enough even when its giving me issues since i dont always need a ton of transfer speed.
The Tailscale docs have information on performance best practices.
And they have at least one video on performance.
I find performance when traveling to be okay and mainly depends on what I am connected to.
At home, I use local network mostly but sometimes forget and days later realize I’m using the Tailnet. There is some difference in speed but not normally that noticeable in (my) real use case. So yes not always the best option, but it’s not hard to switch to the .local net.
Are the devices connecting via their LAN IP (e.g. 192.168.x.x) or WAN IP? You can see with tailscale status:
100.x.x.x fs tagged-devices linux active; direct 192.168.x.x:41641, tx 2148535236 rx 48520952772
If the connection is “direct” but via the WAN IP with hairpin NAT, then the traffic would be going device 1 → switch → router → switch → device 2, which may cause a significant performance impact over device 1 → switch → device 2.
