mirror of
https://github.com/fosrl/pangolin.git
synced 2026-07-11 10:50:35 -05:00
Unable to Connect to Newt #439
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Blucose on GitHub (Jun 15, 2025).
I'm unable to create any docker Newt that connects to my pangolin.
I am able to reach the server with netcat, so ports don't seem to be the problem. The logs for Newt show that it is unable to reach my pangolin server at 100.x, after a successful wireguard handshake. I've tried on multiple machines.
Any help appreciated, happy to provide any logs required.
@jackrosenberg commented on GitHub (Jun 15, 2025):
If it's only the ping that is failing, try rebooting the edge server.
@Blucose commented on GitHub (Jun 15, 2025):
I assume the edge server is the server running newt. I've rebooted and no change. Both on NixOS and Truenas.
@jackrosenberg commented on GitHub (Jun 15, 2025):
Ah thats my bad, I meant to say restart the VPS :). What do the logs say?
@Blucose commented on GitHub (Jun 15, 2025):
the logs for Newt on the Edge (in this case truenas docker compose)
The logs for Pangolin on the VPS
Gerbil on VPS
@jackrosenberg commented on GitHub (Jun 15, 2025):
looks to me like a mismatch between network interfaces:
Internal server is running on http://localhost:3001andError fetching remote config http://pangolin:3001/api/v1/gerbil/get-config.I believe this is a setting somewhere, though i'm not familiar with how TrueNas does this. If this isn't is im afraid this is about the extent of my expertise :)
@Blucose commented on GitHub (Jun 15, 2025):
All of those are on the VPS running Ubuntu no? It's only Newt running on Truenas.
Also those errors seem to eventually go away, I think it's just a delay until pangolin has started fully.
@oschwartz10612 commented on GitHub (Jun 15, 2025):
^^^ yeah I think the get config is fine.
This is almost always a firewall issue when it cant ping. But I know you say you are able to reach the server with netcat. Did you test netcat with UDP port 51820 and get success? Could you triple check port UDP 51820 is open on BOTH the VPS firewall from the provider if there is one and in linux with UFW/iptables/nftables?
You can run a TCP dump when newt connects and look for that 51820 UDP traffic. Let us know if you are getting traffic on that port and it still cant connect.
@Blucose commented on GitHub (Jun 15, 2025):
When I downed the compose on my VPS and listened to UDP 51820 I was able to see the netcat message sent from the edge server.
How do I run a TCP dump?
Appreciating all your help guys!
@oschwartz10612 commented on GitHub (Jun 15, 2025):
Okay thats a pretty good sign then that the firewall is okay. To tcpdump you can do
tcpdump -i any -n udp port 51820Could you double check your docker compose correctly exposes the udp 51820 port? Did you install from the installer?
@Blucose commented on GitHub (Jun 15, 2025):
This is the a snapshot of the output of the TCP dump from the Truenas System while Newt starts up:
...
@cezarq commented on GitHub (Jun 16, 2025):
Same error here:
newt | WARN: 2025/06/16 01:16:24 Ping attempt 172 failed: failed to read ICMP packet: i/o timeout
newt | WARN: 2025/06/16 01:16:24 Ping attempt 63 failed: failed to read ICMP packet: i/o timeout
newt | WARN: 2025/06/16 01:16:24 Ping attempt 480 failed: failed to read ICMP packet: i/o timeout
newt | WARN: 2025/06/16 01:16:26 Ping attempt 479 failed: failed to read ICMP packet: i/o timeout
newt | INFO: 2025/06/16 01:16:26 Pinging 100.89.128.1
@oschwartz10612 commented on GitHub (Jun 16, 2025):
@cezarq could you triple check port UDP 51820 is open on BOTH the VPS firewall from the provider if there is one and in linux with UFW/iptables/nftables?
@Blucose sounds like you did the dump on Truenas running newt and not from the VPS? Could you do it from the VPS to make sure the wg traffic is making it there?
@oschwartz10612 commented on GitHub (Jun 16, 2025):
Also if you use CF you have CF proxy off because it obscures the IP of the VPS and WG traffic cant get through. If you need to use CF proxy you can set the base_endpoint section of the gerbil config file to have the IP of the VPS and it should work for you.
@cezarq commented on GitHub (Jun 16, 2025):
NEWT Client log:
@Blucose commented on GitHub (Jun 16, 2025):
I've got mine working. For me the issue was that the Docker Socket Integration - it wasn't failing gracefully.
When setting up Newt in the UI, if I turn off the toggle for Docker Sock, everything works.
Not sure if there is some bug, or if it is interfering with tailscale or komodo or something..?
@cezarq commented on GitHub (Jun 16, 2025):
@Blucose did you just disable the Docker Socket?
@Blucose commented on GitHub (Jun 16, 2025):
Just did this and it worked. Might investigate later what is causing it to break... but right now just happy it works!
@cezarq commented on GitHub (Jun 16, 2025):
Didnt work for me :-(
Thanks anyway @Blucose
@oschwartz10612 commented on GitHub (Jun 17, 2025):
@Blucose do you have a firewall on the VPS? Who is your provider? Some VPS providers require you to open ports through their UI.
Also double check gerbil is starting correctly but sounds like it is. You could post some log outputs.
Is it possible the network that newt is on is blocking wireguard traffic?
@cezarq commented on GitHub (Jun 17, 2025):
@oschwartz10612, taking advantage of the question, I am using Oracle Cloud (OCI).
@oschwartz10612 commented on GitHub (Jun 17, 2025):
Oh I know some people have had issues with Oracle in the past. Could you take a look at the bottom comments on these two issues and see if it helps?
https://github.com/fosrl/pangolin/issues/818
https://github.com/fosrl/pangolin/issues/889
@cezarq commented on GitHub (Jun 17, 2025):
@oschwartz10612 , no luck! What is weird is that was working 5 days ago perfectly and now I cant go online and I have no idea the reason.
Gerbil log:
Pangolin log:
Traefik log:```
2025-06-16T14:19:52Z INF Traefik version 3.4.1 built on 2025-05-27T12:45:13Z version=3.4.1
2025-06-16T14:19:52Z INF
Stats collection is disabled.
Help us improve Traefik by turning this feature on :)
More details on: https://doc.traefik.io/traefik/contributing/data-collection/
2025-06-16T14:19:52Z INF Loading plugins... plugins=["badger"]
2025-06-16T14:19:52Z INF Plugins loaded. plugins=["badger"]
2025-06-16T14:19:52Z INF Starting provider aggregator *aggregator.ProviderAggregator
2025-06-16T14:19:52Z INF Starting provider *file.Provider
2025-06-16T14:19:52Z INF Starting provider *traefik.Provider
2025-06-16T14:19:52Z INF Starting provider *http.Provider
2025-06-16T14:19:52Z INF Starting provider *acme.ChallengeTLSALPN
2025-06-16T14:19:52Z INF Starting provider *acme.Provider
2025-06-16T14:19:52Z INF Testing certificate renew... acmeCA=https://acme-v02.api.letsencrypt.org/directory providerName=letsencrypt.acme
@cezarq commented on GitHub (Jun 17, 2025):
Pangolin agent:
@cezarq commented on GitHub (Jun 17, 2025):
I dont know if is relevant, but I have a local site configured and is working fine with a service running on the OCI as docker.
@cezarq commented on GitHub (Jun 17, 2025):
I figured out! It was the Cloudflare DNS Proxy. After disabled this feature, back online again.... thanks guys...