Railway•3mo ago

Frontend accessing backend - internal URL?

Hi folks I have a seperate FE and BE service - two different projects. The FE has an env var to locate the backend. Should I set this to the internal URL of the BE service that railway exposes? If so, should I prefix with http/https?

37 Replies

Percy•3mo ago

Project ID: c18400df-5dee-4fe7-bc56-a98db8d475bd

tjhOP•3mo ago

c18400df-5dee-4fe7-bc56-a98db8d475bd

Brody•3mo ago

two different projects, or two different services? please read this page so we can get our terminology correct before going forward - https://docs.railway.app/overview/the-basics

tjhOP•3mo ago

two different services within the same project

Brody•3mo ago

thank you is this a CSR frontend?

tjhOP•3mo ago

NextJS frontend, blend of SSR/CSR some of my build steps hit the backend, when I've changed to the internal URL (with https:// prefix) I get this:

#12 22.31 TypeError: fetch failed

#12 22.31     at node:internal/deps/undici/undici:12502:13

#12 22.31     at process.processTicksAndRejections (node:internal/process/task_queues:95:5)

#12 22.31     at async a (/app/.next/server/.../page.js:1:11099)

#12 22.31     at async u (/app/.next/server/.../page.js:1:10740) {

#12 22.31   digest: '4154157414',

#12 22.31   [cause]: Error: getaddrinfo ENOTFOUND <URL>

#12 22.31       at GetAddrInfoReqWrap.onlookupall [as oncomplete] (node:dns:120:26)

#12 22.31       at GetAddrInfoReqWrap.callbackTrampoline (node:internal/async_hooks:130:17) {

#12 22.31     errno: -3008,

#12 22.31     code: 'ENOTFOUND',

#12 22.31     syscall: 'getaddrinfo',

#12 22.31     hostname: '<NAME>.railway.internal'

#12 22.31   }

#12 22.31 TypeError: fetch failed

#12 22.31     at node:internal/deps/undici/undici:12502:13

#12 22.31     at process.processTicksAndRejections (node:internal/process/task_queues:95:5)

#12 22.31     at async a (/app/.next/server/.../page.js:1:11099)

#12 22.31     at async u (/app/.next/server/.../page.js:1:10740) {

#12 22.31   digest: '4154157414',

#12 22.31   [cause]: Error: getaddrinfo ENOTFOUND <URL>

#12 22.31       at GetAddrInfoReqWrap.onlookupall [as oncomplete] (node:dns:120:26)

#12 22.31       at GetAddrInfoReqWrap.callbackTrampoline (node:internal/async_hooks:130:17) {

#12 22.31     errno: -3008,

#12 22.31     code: 'ENOTFOUND',

#12 22.31     syscall: 'getaddrinfo',

#12 22.31     hostname: '<NAME>.railway.internal'

#12 22.31   }

Brody•3mo ago

the private network is not available during build time. the internal domains may use domains instead of IP addresses, but they need to be treated like a local IP, meaning you use http and a port in the URL. if your frontend calls the backend via SSR during runtime use the private network. if a client's browser calls the backend, use the public url.

tjhOP•3mo ago

Are the IPs static on a hobby plan?

Brody•3mo ago

that is a pro only feature, and it doesnt have anything to do with private networking.

tjhOP•3mo ago

all makes sense, thanks Brody

Brody•3mo ago

let me know if you have any more questions

tjhOP•3mo ago

sorry, one more, when addressing the internal url in SSR, do I use http or https prefix?

Brody•3mo ago

http treat it like your local network, but instead of a ip, it uses a domain

tjhOP•3mo ago

thanks man. Is that something that's obvious to everyone but me, or should it be in docs? (Or, is it already?)

Brody•3mo ago

its not obvious to everyone, i can for sure see the logic in thinking you would use https with a domain name, but it is in the docs! - https://docs.railway.app/guides/private-networking#use-internal-hostname-and-port

tjhOP•3mo ago

thanks again - should have read the manual

Brody•3mo ago

its not always a first thought to dig into the docs, no worries!

tjhOP•3mo ago

ok - have been trying this, on an API call that is definitely SSR I'm calling <my internal url>:8080/<api> from my frontend. My backend (python) has the internal url in ALLOWED_HOSTS I'm getting a

    code: 'ECONNREFUSED',

    syscall: 'connect',

    code: 'ECONNREFUSED',

    syscall: 'connect',

on the frontend

Brody•3mo ago

please provide the actual url you are using

tjhOP•3mo ago

DM'd

Brody•3mo ago

#🛂｜readme #5

tjhOP•3mo ago

you need a way people can pass you variables they don't want exposed would rather not post in public

Brody•3mo ago

I assure you it's not sensitive in the slightest

tjhOP•3mo ago

http://share-a-podcast-be-prod.railway.internal:8080/episode/b71503e3-a130-48a3-9133-9dfd22a05b29/

Brody•3mo ago

it wouldn't be a very private network otherwise haha does your backend listen on 8080?

tjhOP•3mo ago

yes

Brody•3mo ago

gunicorn? unicorn?

tjhOP•3mo ago

verified with a print on launch of my backend - THE PORT IS 8080 gunicorn - started with web: python manage.py migrate && gunicorn <name>.asgi:application -k uvicorn.workers.UvicornWorker --bind 0.0.0.0:$PORT

Brody•3mo ago

remove the bind flag (and it's accompanying host:port)

tjhOP•3mo ago

doing so no joy

Brody•3mo ago

show me your new start command please

tjhOP•3mo ago

web: python manage.py migrate && gunicorn share_a_podcast_be.asgi:application -k uvicorn.workers.UvicornWorker

Brody•3mo ago

what error are you getting now

tjhOP•3mo ago

Same as above Connection refused

Brody•3mo ago

very sorry, i was mistaken, your start command needs to be -

web: python manage.py migrate && gunicorn <name>.asgi:application -k uvicorn.workers.UvicornWorker --bind [::]:$PORT

web: python manage.py migrate && gunicorn <name>.asgi:application -k uvicorn.workers.UvicornWorker --bind [::]:$PORT

tjhOP•3mo ago

works! Next is now giving me warnings about mixed protocols being used (https vs http) Which I’ll look at

Brody•3mo ago

sounds good

Gaming

Programming

Frontend accessing backend - internal URL?