408 error after about a minute of working about server HOT 27 OPEN

After non stop debugging and testing the whole day and night, I (hopefully) found a solution. At least it works locally and has nothing todo with localhost / 127.0.0.1

This simple line in server/TunnelAgent.js makes the magic happen:

     // new socket connection from client for tunneling requests to client
     _onConnection(socket) {
         // no more socket connections allowed
         if (this.connectedSockets >= this.maxTcpSockets) {
             this.debug('no more sockets allowed');
             socket.destroy();
             return false;
         }

+        socket.on("data", () => { /* silence is golden */ });

         socket.once('close', (hadError) => {
             this.debug('closed socket (error: %s)', hadError);
             this.connectedSockets -= 1;

It ensures for some reason (didn't look deeper into it for now) that the connection is getting reestablished after getting dropped. Without this line an important other event (don't know which one for now) is not getting fired which results in broken sockets (9+1 as defined by default).

Will try to implement these changes on my live server as soon as I got some sleep and hopefully it works there too.

Example:
Without this "fix":

With this "fix":

Tested on my VPS with success 🎉
Works with all versions of Node (16,18,20)

from server.

Exactly, 18 | 18/20 | 18 | 127.0.0.1/locahost is what we wanna get up and running. In fact - every time a user uses Node v18 or higher, localtunnel stops working after about 1 minute. I guess the socket connection just drops or something.

I already tried to update all dependencies to its latest version - with the same result - getting HTTP ERROR 408 after some time and dropping connection when max socket connection is reached.

Also thought about using socket.io for connection but it's not that easy to make it work. Reason therefore is, that this repository uses socket.io (https://github.com/BenMcH/node-grok) and it simply works! But without a subdomain like localtunnel :(

Will try to replace koa with express like you mentioned, maybe this gives me some other responses or ideas.

Were you able to reproduce the error I mentioned with the 408 http error?

from server.

@tiagopazhs quick update after debugging the hell out of locatunnel and your repository of localTunnel-OpenSource:

I debugged the problem down to the net.connect since I figured out, that the local connection just drops with http error 408. The local connection which maps the local application to the client service.

As I thought, the reason is this: nodejs/node#43776

I used your localTunnel-OpenSource to debug since socket.io and express just gives me more details when running it with DEBUG=*.

Now I'll try to implement the changes I've made into localtunnel.. let's see

from server.

from server.

It’s is your first time hosting at AWS? Did you run local tunnel local and got the same error ?
…
On Wed, 23 Aug 2023 at 07:22 rawvie-tnt @.> wrote: I am hosting the local tunnel server in an AWS ec2 instance, after setting up everything when, I try to connect from the client, the tunnel works for about a minute then shows HTTP ERROR 408. site.conf from nginx/conf.d: proxy_http_version 1.1; server { listen 0.0.0.0:80; location / { proxy_pass http://127.0.0.1:3000/; proxy_set_header X-Real-IP $remote_addr; proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for; proxy_set_header Host $http_host; proxy_set_header X-Forwarded-Proto http; proxy_set_header X-NginX-Proxy true; proxy_set_header Upgrade $http_upgrade; proxy_set_header Connection $connection_upgrade; proxy_redirect off; } } nginx.conf: user nginx; worker_processes 1; error_log /var/log/nginx/error.log warn; pid /var/run/nginx.pid; events { worker_connections 1024; } http { include /etc/nginx/mime.types; default_type application/octet-stream; log_format main '$remote_addr - $remote_user [$time_local] "$request" ' '$status $body_bytes_sent "$http_referer" ' '"$http_user_agent" "$http_x_forwarded_for"'; access_log off; sendfile on; #tcp_nopush on; gzip on; gzip_min_length 1000; include /etc/nginx/conf.d/.conf; include /etc/nginx/sites/; } why does it work just for about a minute. I tried it from both CLI and API. — Reply to this email directly, view it on GitHub <#203>, or unsubscribe https://github.com/notifications/unsubscribe-auth/ATONHPRWSH2C55RIEMEQJ6LXWXKYFANCNFSM6AAAAAA33HT5DA . You are receiving this because you are subscribed to this thread.Message ID: @.**>

yes is its is my first time running local tunnel in AWS, but i have tried with multiple different ec2's with same result, when connecting with the default host server it works as expected, the issue occurs when I try to connect with my server

from server.

As it is yout first time running local tunnel server. Can you run it locals first and let me know if you get a similar issue?

Follow this 2 commits to run local: 

Fix esm modules

It will help you to setup the project because there are some differences between node versions.

https://github.com/tiagopazhs/localtunnelServer-raise/commit/425c9733393a13dedd32a363c325a076d4006cbf

Run server in your machine

If you debug your application you will got this problem because localtunnel uses an especific lib that don't handle with local urls.

https://github.com/tiagopazhs/localtunnelServer-raise/commit/518214b9163b4cc80b50f2d9d36686904179f344

from server.

Facing the same issue. After around one minute the assigned domain just returns error 408 on self hosted server. When I use the default hosted one, it works. I'd like to fix this but I don't really know where to look at. Server? Client? But I guess it's a server problem.

from server.

@creazy231 can you share error logs ?
Did you already clone localtunnel/client and clone localtunnel/server and running local in your machine ?

from server.

@tiagopazhs Answering here since I saw tiagopazhs/localTunnel-OpenSource#1 as an alternative.

Yes, I already setup everything and before Node v18+ everything worked fine. Thought it could be the server so I also updated to Node v18 but since then I'm getting a http error code 408 in the browser with Node v16 and Node v18 after exactly one minute (+/- 10 seconds)

I guess you should be able to reproduce using npx lt --host http://lt.boltapi.com --local-host 127.0.0.1 --port 1337 for example. localtunnel/server is hosted on lt.boltapi.com using Node v18 + PM2 with modifications you mentioned here using this commit https://github.com/tiagopazhs/localtunnelServer-raise/commit/425c9733393a13dedd32a363c325a076d4006cbf

also tried to use the docker version on the server - same result.

The only thing works is localtunnel.me - the default host - so I guess there had to be some changes on their side they're using.

from server.

@creazy231 so, if you share your repository, I can hel you.
I ran it using node 20.7.0 with out problems.
So, let me know if you follow this step by step.

Fix esm modules

Step 1:

https://github.com/tiagopazhs/localtunnelServer-raise/commit/425c9733393a13dedd32a363c325a076d4006cbf

Step 2:

"start": "node ./bin/server.js --port 3043",

Step 3:

npm i

npm run start

Step 4 (optional):

As you debug the application in your machine you can face an issue about the period od grace. For safety reasons you have just 1 second to connect. And if you connect out of this time. You'll be redirected to the landing page: https://localtunnel.github.io/www/
Update the period of grace to degub the application in your client.js file:

Now you have a localtunnel server running in your own machine :

Run the application that you want to share with the world

I recomend you run a hello world server like this:

const http = require('http');

const server = http.createServer((req, res) => {
  res.writeHead(200, { 'Content-Type': 'text/plain' });
  res.end('Hello, Tobias!\n');
});

const port = process.env.PORT || 7000;

server.listen(port, () => {
  console.log(`Server running at http://localhost:${port}/`);
});

Connect the client to your server

Step 1:

https://github.com/localtunnel/localtunnel

Step 2:

node bin/lt.js --port 7000 --host http://localhost:3043 --subdomain tobias-28

Step 3 !IMPORTANT!:

    function GetClientIdFromHostname(hostname) {
        hostname = hostname.replace(':3004', '.com.br')
        return myTldjs.getSubdomain(hostname);
    }

from server.

@tiagopazhs thanks for this awesome setup guide ❤️
I'll follow it with exactly the same settings using npm instead of yarn and also using Node v20 instead of Node 18 and will come back as soon as I'm done. Cross my fingers to be able to make it work

from server.

@tiagopazhs after exact same setup it works for some time and page reloads.. until this happens upon reload:

Client:

  ...
  localtunnel:client establishing tunnel http://localhost:4000 <> lt.boltapi.com:36461 +0ms
  localtunnel:client tunnel open [total: 10] +31ms
  localtunnel:client connecting locally to http://localhost:4000 +64ms
  localtunnel:client connected locally +5ms
  localtunnel:client local connection closed [false] +44s
  localtunnel:client local connection closed [false] +0ms
  localtunnel:client local connection closed [false] +0ms
  localtunnel:client local connection closed [false] +30s
MacBook-Pro-von-Tobias-2:test thiele$

after local connection closed on client side, Server logs:

-- Start getting HTTP ERROR 408 in Browser
  lt:Client[tobias-28] > / +2m
  lt:TunnelAgent[tobias-28] create connection +2m
  lt:TunnelAgent[tobias-28] socket given +1ms
  lt:Client[tobias-28] < / +4ms
  lt:TunnelAgent[tobias-28] closed socket (error: false) +4ms
  lt:TunnelAgent[tobias-28] connected sockets: 3 +0ms
-- Refresh page -> getting HTTP ERROR 408 in Browser
  lt:Client[tobias-28] > / +204ms
  lt:TunnelAgent[tobias-28] create connection +202ms
  lt:TunnelAgent[tobias-28] socket given +1ms
  lt:Client[tobias-28] < / +3ms
  lt:TunnelAgent[tobias-28] closed socket (error: false) +5ms
  lt:TunnelAgent[tobias-28] connected sockets: 2 +0ms
-- Refresh page -> getting HTTP ERROR 408 in Browser
  lt:Client[tobias-28] > / +104ms
  lt:TunnelAgent[tobias-28] create connection +100ms
  lt:TunnelAgent[tobias-28] socket given +1ms
  lt:Client[tobias-28] < / +1ms
  lt:TunnelAgent[tobias-28] closed socket (error: false) +1ms
  lt:TunnelAgent[tobias-28] connected sockets: 1 +0ms
-- Refresh page -> getting HTTP ERROR 408 in Browser
  lt:Client[tobias-28] > / +129ms
  lt:TunnelAgent[tobias-28] create connection +129ms
  lt:TunnelAgent[tobias-28] socket given +0ms
  lt:Client[tobias-28] < / +2ms
-- Refresh page -> getting HTTP ERROR 408 in Browser
  lt:TunnelAgent[tobias-28] closed socket (error: false) +3ms
  lt:TunnelAgent[tobias-28] connected sockets: 0 +0ms
  lt:TunnelAgent[tobias-28] all sockets disconnected +0ms
  lt:Client[tobias-28] client offline tobias-28 +2ms
  lt:Client[tobias-28] > / +113ms
  lt:TunnelAgent[tobias-28] create connection +114ms
  lt:TunnelAgent[tobias-28] waiting connected: 0 +0ms
  lt:TunnelAgent[tobias-28] waiting available: 0 +1ms
  lt:ClientManager removing client: tobias-28 +2m
  lt:TunnelAgent[tobias-28] closed tcp socket +887ms
  lt:TunnelAgent[tobias-28] closed tcp socket +0ms
-- Refresh page -> getting HTTP ERROR 404 in Browser

Maybe you can reproduce this by just idling in the browser for some time without refreshing etc. and then refresh page after around one or two minutes.. at least using lt.boltapi.com as host you should face the same problem.
It's weird because when I constantly refresh the page for like once every second, the connection keeps stable.

Using Ubuntu 22.04, nginx with LetsEncrypt wildcard cert, Cloudflare for DNS without Proxy.

I just don't know what to do anymore at this point to make it work.. even changing back to Node v16 on server and client doesn't change anything.

Edit:
after all that struggle, I finally cloned and setup the whole project locally as you described - same result. After idling for some time - HTTP ERROR 408
ever additional refresh of the page results in an additional socket connection. When it reaches 10 (default max) it drops the connection and closes the tunnel.

Edit:
Running the whole project on localhost using Node v16 - it works

Some Testing:

so it points out, that the process has to run on Node v16 to make it work. The problem seems to be not related to localtunnel itself but changes made from Node v16 to Node v18. Maybe nodejs/node#40537 is related to this? But even changing everything to 127.0.0.1 results in the same error

from server.

This is a very good analysis! So I didn't know that.

So, I don't know what is releated to this problem.

Maybe to update some dependencies. In the past I used Express instead Koa and it works well but I can't say that this is the problem.

So, using this bellow scenario everything works fine? Can you see your local application running in other machine ?
16 | 20 | 16 | ✅ | 127.0.0.1

In this moment our main goal is to run with this scenario, rigth ? I will try to solve that.

from server.

Sorry for my latest answer. Really ? Congratiulations 🎉.
You made an incredible work and now we have a localtunnel able to revive.

from server.

@creazy231 when I use nove v18 the client is don't found the server.
Is there another thing to do ?
ps: I am already using the magic line: socket.on("data", () => { /* silence is golden */ });

from server.

@tiagopazhs connecting with node bin/lt.js --port 4000 --host http://localhost:3043 on local and using lt --host http://lt.boltapi.com --local-host 127.0.0.1 --port 3000 --subdomain my-sub-domain connecting to my VPS LocalTunnel. Didn't change anything else on client side.

I guess the important part is the --local-host 127.0.0.1 using Node v18+ because of the default dns change since Node v17. Also make sure your local apps also running on 127.0.0.1

from server.

Sorry for the late answer. For me it is not working yet:

My server is running on 18 version:

But when I use version 18 in my client. It can't find anything:

from server.

I used this fork and everything worked.

from server.

@rawvie-tnt thanks for this awesome way to use localtunnel. The fork has interesting features and ways to make localtunnel better.

So, about the previous error that I mentioned: #203 (comment)
I am trying to get the answer for why it is happening. Do you know how the author of the fork did to solve this bug in node 18 ?

from server.

@tiagopazhs I followed your instructions but am facing the below issue. The URL is not coming properly. Could you please help me? I have not been able to figure out the problem.

I am using this fork mentioned by @rawvie-tnt

#203 (comment)

your url is: http://tobias-28.undefined

from server.

@tiagopazhs I followed your instructions but am facing the below issue. The URL is not coming properly. Could you please help me? I have not been able to figure out the problem.

I am using this fork mentioned by @rawvie-tnt

#203 (comment)

your url is: http://tobias-28.undefined

Try starting localtunnel server using --domain <your_domain_name>. Then when you connect the client with subdomain, the client url is shown correctly. It works for me for this fork of localtunnel server.

I was able to connect node v18 client running nodev18 process to node v18 server properly. Hope this helps !

from server.

@SlitheryMender Thanks for your reply! May I know where did you deploy the localtunnel-server to get the domain? I am trying to deploy it to https://onrender.com/ but it is not working.

from server.

@rawvie-tnt I am also using the EC2 instance to run the server. Only HTTP URLs are functional. The issue I am encountering is the inability to add HTTPS support. I have attempted to point the DNS records to the ALB domain by attaching an HTTPS listener with the ACM certificate, but it is not working. Did you add HTTPS support?

from server.

guess your server setup is kinda wrong. I'm running a wildcard cert using certbot and everything works

from server.

@creazy231 Yes, but I'm not sure where the problem lies. Are you also utilizing the EC2, ALB, and Route 53 services to host the server code?

from server.

nope.. running on a 3€/mo hetzner cloud server because I don't like the complexity of AWS for simple braindead projects

from server.

It looks like the issue has been around for a year. What I can tell is that this is not a server (only) problem: on the same machine, using a Node 16 client (via Electron) created a stable tunnel, but using a Node 18 client (via Visual Studio Code Extension Host) create a tunnel that broken at some point, with 408 error code.

I will try this fork tomorrow and let you guys know if it works. And I wonder what exactly is modified in the fork? I took a quick glance at the diff, and the only thing I saw was code formatting...

from server.