cb5aa03d-c72c-426c-8dd7-f109b7f03b5d

there's a wonderful guide here

but yes it's good your apps are listening on the automatically assigned PORT variable, so that means getting your apps to listen on a static port for use with the private network is as simple as setting a PORT service variable

Yeah, found the guide, guessed it was the case that I would have to define a port instead 😄 No worries.

what kind of frontend is this?

Sveltekit

server side rendered?

🤷‍♂️ I'm a frontend noob, much more comfortable kneedeep in AWS infra and Python 😄

I manage a bunch of Svelte devs so figured I should learn it a bit haha

will the calls to the backend be made from the server side of your frontend app?

Yes, it already is, you can't see the requests in devtools for example

perfect then you will want a service variable on the frontend service like this

BACKEND_URL=http://${{.RAILWAY_PRIVATE_DOMAIN}}:${{.PORT}}

then use that variable in your frontend's code when you make the fetch call

Ohlala.

Don't suppose there's a page with all these magic env vars?

oh yes my bad

Sweet. That snippet you sent would be super useful on the Private Networking guide, makes much more sense now 😄

Annoyingly I don't seem to have the Solution App available

I shall make that happen

I have another question on the community page, feel bad duplicating it here also 🥲 Buuuuut if you're around 👀

solution app?

I turned that off because I had too many people marking the complete wrong message as the answer

Ah 😛

but I haven't turned off the message lol

No worries haha

I'll be honest I'm happy you posted here too, and I'd prefer you do, because as you said discord is easier for conversational things.

the forums is lacking a conversational aspect and that's why I don't like using them, but the cheerleader for the forums knows that too!

Righto. I'll make a dupe here then 😅

sounds good!

Back on this one 😅 So I put:

BACKEND_URL=http://${{backend.RAILWAY_PRIVATE_DOMAIN}}:${{backend.PORT}}

This resolves but if I put Https or http, it fails to connect, the more concerning part is the PORT resolves to not what the backend service is actually listening on:

errno: -111,

code: 'ECONNREFUSED',

syscall: 'connect',

address: 'fd12:2030:e370::8f:bc37:5946',

Backend logs:

[INFO] Running on http://[::]:6642

Am I using the right variables?

Hmmm okay, in the dashboard it resolves ${{backend.RAILWAY_PRIVATE_DOMAIN}} as backend.railway.internal however when that's injected into my service, it's actually the ipv6 address.

If I then use [] for fetch like recommended for ipv6 addresses, it resolves to:

Safe to say I'm scuppered 😛

BACKEND_URL=http://[${{backend.RAILWAY_PRIVATE_DOMAIN}}]:${{backend.PORT}}

This is what I have right now, Railway setup the port for backend, so I assume I'll be able to access that value.

This resolves to 'http://[backend.railway.internal]:'

If I remove the [] from my ENV (Even though that's not what I should do), it resolves to the actual ipv6 address.

Either way I do it, PORT Is empty anyway 🥲

Okay, so a little progress:

BACKEND_URL=http://${{backend.RAILWAY_PRIVATE_DOMAIN}}:6642

If I hardcode the port to what I see in my backends logs (from the port Railway generated), I get a connection

So I guess the real question is - How the heck do I get that port without log diving 😄

(I'd rather not pin it as it just another bit of config to remember)

pinning the port by setting a PORT service variable is perfectly fine as long as your app is still listening on the PORT environment variable, and that's exactly what you need to do to get that reference variable to render correctly.

it's also absolutely necessary because without pinning the port, railway will assign a random port (via the PORT environment variable) on every deployment that it expects your app to listen on, you can't reference this magic variable and even of you could it would just makes things worse because its random

also I have never seen the square brackets be necessary, so try not using the brackets

Yeah, it's just strange that if I set the value, I can refer to it, but if Railway does, I can't 😂

Alright, I shall just pin it!

Is that something I can define in railway.json? Or dashboard only?

yeah I get ya, but it's a good thing you can't reference the automatically generated PORT, as the value would become stale very quickly anyway, it's random and only available to the deployment at runtime.

has to be defined as a service variable in the dashboard

Yeah that's fair enough, just somewhat invalidates the 'magic' 😂

Anyway, I'll switch to that, thanks once again Brody! You the MVP

its now in the docs!

Hi folks,

I'm having trouble using private networks
I read all the content above, including the links, followed the tutorial in the last link sent above, but I was unsuccessful.

I have two backend services running on railway, where currently one communicates with the other via the public network.

I would like to leave this second service with the public route disabled, and use the private network to communicate with it.

The two services have the same project name and environment, as I understood from the document above.
The service that will receive requests via the private route is configured to listen on IPv6.

I tested calling the url numerous times using ###.railway.internal and port, but it always returns "getaddrinfo ENOTFOUND"

I don't know what else I can try, does anyone suggest anything?

Both projects are in nestJs with typescript.

nixpacks or dockerfile based deployments?

Nixpacks

at what point in your apps lifecycle are you trying to call the other service?

The first service, with public routes, is an API gateway that, depending on the requested route, makes a call to this second service.
This call I would like to make via private route internally.

does your service make a call to the other service within the first 3 seconds of itself starting?

no, never

screenshot of your project please

what do you want i screen shot exactly?

of your railway project

the second service

where is the other service

the first, that uses the second

these are separate projects

they are different reppositories

is there any how deploy toggether?

ok, i got it

i just saw the "new" button

you can't communicate to services in different projects

wouldn't be very private if you could

yeah, i undertand it
i thought as i set the project name as the same, it would work as one

gotcha

i use railway for almost 2 years the wrong way 🥲
I'd never seen that "new" button

i have a lot to do now.. re-managing the services of my projects

yeah you aren't the first

and until railway stops allowing users to deploy services from the dashboard, you will not be the last

now it makes more sense why it has to be in the same project AND enviroment

indeed

thank you
its already working

awesome

Hello @Brody , I have a question

I want my front to connect to my backend using Private Network, but I haven't been able to do it… I checked this post but nothing helps… any ideas?

I have this as follows:

And my code is this:

export const getStrapiURL = () => {

    if (process.env.NODE_ENV === 'production') {
        return 'https://espectrosoft-backend.railway.internal'
    }

    if (process.env.NODE_ENV === 'development') {
        return 'http://localhost:1337';
    }
}

a client side rendered frontend app can not use the internal domain of the backend, you need to use the public domain.

it wouldnt be a very private network if anyone's browser could make requests to it.

I'm not exactly using client side, I'm doing it with Nextjs and I make these requests using server side

please read this then -

I tried to do that but in my variables I can't find the port

you need to set it then

And it can be any port or is there any guide regarding that?

8080 is a good port

make sure your app is listening on the $PORT environment variable too

Got it, I'll try it