Very slow speeds remotely #1644

@chemicalsam commented on GitHub (Feb 22, 2025):

It seems better using a relatives wifi network, about 7mbps. More usable than before. Idk if tmobile is somehow throttling it?

@chemicalsam commented on GitHub (Feb 22, 2025): It seems better using a relatives wifi network, about 7mbps. More usable than before. Idk if tmobile is somehow throttling it?

saavagebueno commented

@Chanceium commented on GitHub (Feb 24, 2025):

I can confirm that I have noticed NetBird is quite slow. I have the exact same setup using Headscale self-hosted, and with a 100 Mbps line speed, I get around 85 Mbps back after the overhead with Tailscale. However, with NetBird, it’s more like around 10 Mbps. They are both using WireGuard though correct?

@Chanceium commented on GitHub (Feb 24, 2025): I can confirm that I have noticed NetBird is quite slow. I have the exact same setup using Headscale self-hosted, and with a 100 Mbps line speed, I get around 85 Mbps back after the overhead with Tailscale. However, with NetBird, it’s more like around 10 Mbps. They are both using WireGuard though correct?

saavagebueno commented

@levindecaro commented on GitHub (Mar 16, 2025):

I have same experience too, when relay is selected, the transfer speed drop from 40Mbps to 1.4Mbps, and often dropping packet while transferring large file.

I analysed all logs and debug output, it looks work as expected. I than use tcpdump on the relay server, I could say the transfer rate are quite low. Seems like something is throttling I don't know.

lastly, I was frustrated to figure out why the same client same network sometime cannot negotiate with P2P, I was confirm the client side NAT-PMP is mapping 51820 towards the source client, and i do "nc" on the counter peer that indicating the client 51820 port is reachable, however it still fallback into relay mode.

@mlsmaycon do you have any idea how to troubleshoot this case.

@levindecaro commented on GitHub (Mar 16, 2025): I have same experience too, when relay is selected, the transfer speed drop from 40Mbps to 1.4Mbps, and often dropping packet while transferring large file. I analysed all logs and debug output, it looks work as expected. I than use tcpdump on the relay server, I could say the transfer rate are quite low. Seems like something is throttling I don't know. lastly, I was frustrated to figure out why the same client same network sometime cannot negotiate with P2P, I was confirm the client side NAT-PMP is mapping 51820 towards the source client, and i do "nc" on the counter peer that indicating the client 51820 port is reachable, however it still fallback into relay mode. @mlsmaycon do you have any idea how to troubleshoot this case.

saavagebueno commented

@mlsmaycon commented on GitHub (Mar 17, 2025):

Hello @levindecaro , can you please share the output file from netbird debug bundle -S?

@mlsmaycon commented on GitHub (Mar 17, 2025): Hello @levindecaro , can you please share the output file from `netbird debug bundle -S`?

saavagebueno commented

netbird.debug.417375207.zip

@levindecaro commented on GitHub (Mar 17, 2025):

@mlsmaycon Thank you for your time to investigate the problem. Today I solve packet dropping issue because the routing peer getting DNS 30% timeout per connection, without "netbird up" no DNS issue at all. Once I disable DNS management for routing peer groups. No more packet dropping. And the speed went to to 20Mbps compared with 40Mbps(P2P), it become more usable now.

The remaining uncertain issue is why P2P connection aren't behave consistently on the same client/network.

Besides, do netbird designed to leverage UPnP/NAT-PMP? I can't found a reliable source to discuss that, but people mention about it.

Cheers.

@levindecaro commented on GitHub (Mar 17, 2025): [netbird.debug.417375207.zip](https://github.com/user-attachments/files/19290941/netbird.debug.417375207.zip) @mlsmaycon Thank you for your time to investigate the problem. Today I solve packet dropping issue because the routing peer getting DNS 30% timeout per connection, without "netbird up" no DNS issue at all. Once I disable DNS management for routing peer groups. No more packet dropping. And the speed went to to 20Mbps compared with 40Mbps(P2P), it become more usable now. The remaining uncertain issue is why P2P connection aren't behave consistently on the same client/network. Besides, do netbird designed to leverage UPnP/NAT-PMP? I can't found a reliable source to discuss that, but people mention about it. Cheers.

saavagebueno commented

@Nordlicht-13 commented on GitHub (May 6, 2025):

I also have a slowdown on the local network.
Windows PCs connecting to a ubuntu server with smb-shares on it.
Sometimes it takes a while (some seconds) to change between folders in a explorer.
And moving E-Mails on Windows E-Mail-Client from one IMAP-folder to another on the linux server
sometimes takes also a while or it's not responding any more.
Tailscale has no problem like this on the same PC.

@Nordlicht-13 commented on GitHub (May 6, 2025): I also have a slowdown on the local network. Windows PCs connecting to a ubuntu server with smb-shares on it. Sometimes it takes a while (some seconds) to change between folders in a explorer. And moving E-Mails on Windows E-Mail-Client from one IMAP-folder to another on the linux server sometimes takes also a while or it's not responding any more. Tailscale has no problem like this on the same PC.

saavagebueno commented

@arthur-lbchn commented on GitHub (Jul 11, 2025):

Hi. I'm considering replace Headscale self-hosted with Netbird self-hosted, however, it comes that its almost twice slow. I was testing using the same network and physically same exit nodes on both:

Netbird

Tailscale

@arthur-lbchn commented on GitHub (Jul 11, 2025): Hi. I'm considering replace Headscale self-hosted with Netbird self-hosted, however, it comes that its almost twice slow. I was testing using the same network and physically same exit nodes on both: Netbird <img width="444" height="600" alt="Image" src="https://github.com/user-attachments/assets/edc47367-059c-42e3-ac54-fa9634e5ca56" /> Tailscale <img width="444" height="600" alt="Image" src="https://github.com/user-attachments/assets/df9c5330-1473-4d7c-8661-38e1eac50d47" />

saavagebueno commented

@Nordlicht-13 commented on GitHub (Jul 11, 2025):

Hi. I'm considering replace Headscale self-hosted with Netbird self-hosted, however, it comes that its almost twice slow. I was testing using the same network and physically same exit nodes on both:

Netbird
Tailscale

Can you do a speedtest after disconnecting from the Birdnet and connecting again.

@Nordlicht-13 commented on GitHub (Jul 11, 2025): > Hi. I'm considering replace Headscale self-hosted with Netbird self-hosted, however, it comes that its almost twice slow. I was testing using the same network and physically same exit nodes on both: > > Netbird > > <img alt="Image" width="444" height="600" src="https://private-user-images.githubusercontent.com/9682681/465268834-edc47367-059c-42e3-ac54-fa9634e5ca56.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.D86OZ9HpgKpXfRO2mFxoCPTKcA_t6BHfN27SNwPKgs0"> > Tailscale > > <img alt="Image" width="444" height="600" src="https://private-user-images.githubusercontent.com/9682681/465268884-df9c5330-1473-4d7c-8661-38e1eac50d47.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.H01VOj7v8Os8RlyliBg0UTNUkBk8iOSyqvxELnK6-nw"> Can you do a speedtest after disconnecting from the Birdnet and connecting again.

saavagebueno commented

@arthur-lbchn commented on GitHub (Jul 11, 2025):

Sure, here we go

After disconnecting:

Connecting back to Netbird:

@arthur-lbchn commented on GitHub (Jul 11, 2025): Sure, here we go After disconnecting: <img width="444" height="750" alt="Image" src="https://github.com/user-attachments/assets/c527fecf-c01d-48f5-9643-0f83f75a7532" /> Connecting back to Netbird: <img width="444" height="750" alt="Image" src="https://github.com/user-attachments/assets/5756afd8-22e5-4038-aedd-fdc1e5ad4b30" />

saavagebueno commented

@Nordlicht-13 commented on GitHub (Jul 11, 2025):

Strange, I have normal Internetspeed

Connected to Birdnet:
Download: 298.88 Mbps
Upload: 314.79 Mbps
Ping ms 1 17 2

Disconnected from Birdnet:
Download: 298.91 Mbps
Upload: 315.31 Mbps
Ping ms 1 15 2

I just have speed problem sometimes to SMB-Shares.
Then the seed goes down to around 15 MB/sec.
Disconnecting from Birdnet and reconnecting to the Birdnet again brings back normal seed aroud 115 MB/sec.

@Nordlicht-13 commented on GitHub (Jul 11, 2025): Strange, I have normal Internetspeed Connected to Birdnet: Download: 298.88 Mbps Upload: 314.79 Mbps Ping ms 1 17 2 Disconnected from Birdnet: Download: 298.91 Mbps Upload: 315.31 Mbps Ping ms 1 15 2 I just have speed problem sometimes to SMB-Shares. Then the seed goes down to around 15 MB/sec. Disconnecting from Birdnet and reconnecting to the Birdnet again brings back normal seed aroud 115 MB/sec.

saavagebueno commented

@nazarewk commented on GitHub (Jul 11, 2025):

Are you connected to those directly (P2P) or through Relays? Relays might introduce slowdowns, especially for extremely latency-sensitive SMB protocol (nothing we can do about it except increasing chances of P2P).

@nazarewk commented on GitHub (Jul 11, 2025): Are you connected to those directly (`P2P`) or through Relays? Relays might introduce slowdowns, especially for extremely latency-sensitive SMB protocol (nothing we can do about it except increasing chances of P2P).

saavagebueno commented

@Nordlicht-13 commented on GitHub (Jul 11, 2025):

netbird status -d tells me that truenas is connected through relay.
But disconnecting and connecting from netbird and it's still relay, but with 115 MB/sec.
What can I do to force a P2P connection?

@Nordlicht-13 commented on GitHub (Jul 11, 2025): `netbird status -d` tells me that truenas is connected through relay. But disconnecting and connecting from netbird and it's still relay, but with 115 MB/sec. What can I do to force a P2P connection?

saavagebueno commented

@nazarewk commented on GitHub (Jul 11, 2025):

netbird status -d tells me that truenas is connected through relay. But disconnecting and connecting from netbird and it's still relay, but with 115 MB/sec. What can I do to force a P2P connection?

@Nordlicht-13 could you provide me with a debug bundle after setting a few environment variables and waiting for the slowdown to happen?

Not sure which system are you running, but I want NB_LOG_LEVEL=trace & NB_LOG_MAX_SIZE_MB=200 (to have 200MB of logs - possibly a few hours on TRACE level), you can find instructions for Windows at https://docs.netbird.io/how-to/troubleshooting-client#on-windows
Running 0.50.2 (currently latest) or later you can run netbird debug bundle -C5 -SU to cover as many logs as possible.
If applicable please mention which route/peer you're accessing while observing the slowdown.

@nazarewk commented on GitHub (Jul 11, 2025): > `netbird status -d` tells me that truenas is connected through relay. But disconnecting and connecting from netbird and it's still relay, but with 115 MB/sec. What can I do to force a P2P connection? @Nordlicht-13 could you provide me with a debug bundle after setting a few environment variables and waiting for the slowdown to happen? Not sure which system are you running, but I want `NB_LOG_LEVEL=trace` & `NB_LOG_MAX_SIZE_MB=200` (to have 200MB of logs - possibly a few hours on TRACE level), you can find instructions for Windows at https://docs.netbird.io/how-to/troubleshooting-client#on-windows Running `0.50.2` (currently latest) or later you can run `netbird debug bundle -C5 -SU` to cover as many logs as possible. If applicable please mention which route/peer you're accessing while observing the slowdown.

saavagebueno commented

@Nordlicht-13 commented on GitHub (Jul 11, 2025):

Yes, it happens on a Windows 11 system.
Not really sure what command I should run

Just like this:

[Environment]::SetEnvironmentVariable("NB_LOG_LEVEL", "debug", "Machine")
netbird service restart

Where do I add the NB_LOG_LEVEL=trace and NB_LOG_MAX_SIZE_MB=200?
Do I have to put in for Machine the name of the Windows-Peer?

Or do I have to do this in the Docker compose file on my TrueNAS?
All Peers are on the latest 50.2

@Nordlicht-13 commented on GitHub (Jul 11, 2025): Yes, it happens on a Windows 11 system. Not really sure what command I should run Just like this: ``` [Environment]::SetEnvironmentVariable("NB_LOG_LEVEL", "debug", "Machine") netbird service restart ``` Where do I add the `NB_LOG_LEVEL=trace` and `NB_LOG_MAX_SIZE_MB=200`? Do I have to put in for `Machine` the name of the Windows-Peer? Or do I have to do this in the Docker compose file on my TrueNAS? All Peers are on the latest 50.2

saavagebueno commented

@nazarewk commented on GitHub (Jul 15, 2025):

@Nordlicht-13 you can set the NB_LOG_MAX_SIZE_MB the same way you just did with the NB_LOG_LEVEL for the Windows system.
docker-compose sounds like the right place to add envvars to for your TrueNAS.

It might help or even prove critical to gather bundles & logs from both ends (including your TrueNAS) if you can do that.

@nazarewk commented on GitHub (Jul 15, 2025): @Nordlicht-13 you can set the `NB_LOG_MAX_SIZE_MB` the same way you just did with the `NB_LOG_LEVEL` for the Windows system. `docker-compose` sounds like the right place to add envvars to for your TrueNAS. It might help or even prove critical to gather bundles & logs from both ends (including your TrueNAS) if you can do that.

saavagebueno commented

@riggs498 commented on GitHub (Jul 28, 2025):

I wanted to jump in and say I am having the same issue. My Zerotier connection is way faster that I am getting with NetBird. But I want to really use NetBird vs Zerotier. Since it is all self hosted. My ping times are between 125 and 245. I do a speed test and it is 3.70 down to 9.15 Upload.

If I connect to my local network and then connect to netbird. I still have good speed to my home assistant server.
Good ping also.

As soon as I hit remote. Cell network or work internet. I lose all speed and ping goes to what I stated before.

My server is hosted in the cloud. All those ports are open. There are no ports at home open for this. Home is an exit node. should I open ports at home?

@riggs498 commented on GitHub (Jul 28, 2025): I wanted to jump in and say I am having the same issue. My Zerotier connection is way faster that I am getting with NetBird. But I want to really use NetBird vs Zerotier. Since it is all self hosted. My ping times are between 125 and 245. I do a speed test and it is 3.70 down to 9.15 Upload. If I connect to my local network and then connect to netbird. I still have good speed to my home assistant server. Good ping also. As soon as I hit remote. Cell network or work internet. I lose all speed and ping goes to what I stated before. My server is hosted in the cloud. All those ports are open. There are no ports at home open for this. Home is an exit node. should I open ports at home?

saavagebueno commented