AdonisJS Framework
Replies: 9 comments 55 replies
|
No, we have no plan doing that. You are doing it wrong if you want to run a long-running process in Lambda, what you need is PaaS, not FaaS. |
|
I totally know how it works.
As you said, it is not designed to run a long-running process. I don't care if it works or not. Also, PHP is not a long-running process, so it is a great candidate for lambda. I also don't care about spending 0$, 0.20$, 5$, or even 20$ per month for a server. Stop making things complicated for no reason. |
|
You're not totally right : Laravel with OCTANE is a long running process... and that's how Laravel Vapor run php projects. Having to manage servers and scale them up/down is what I see complicated for no reasons... As a freelance dev, I still smile when I hear my customers telling me "we're paying 200$ a month for our server, we have 4 users on it... how can you tell me it could cost only 0.20$ ??"... |
|
Vapor does not run Laravel Octane by default. It is an opt-in option you can do in any Laravel application deployed anywhere.
Cool, then you are free to spend time on it but do not expect us to research into that space for AdonisJS. |
|
Would you approve a community provided package that would allow to run Adonis on serverless architectures ? |
|
Yes, of course. Feel free to create whatever package you want! We will not stop you from trying. 😄 |
|
Yeah really not something that belongs here. We don't have bandwidth to research into the new space. |
|
@tzsk something you can look for and use is to use only AdonisJS http server |
|
Yup. You are exactly right. I did dig around a bit. There are references of Node Specific The NODE runtime is tightly coupled with the framework, If we'd want to use it we will have to maintain a forked version of the framework. If we could extract Node Runtime objects Hypothetically, and accept the Things required from the node runtime like 'https' to a different object and accept the implementation on the bootstrap of Then we can write an express adapter to run in AWS Lambda. And yes, we can cache the Bootstrapped Application Instance in memory inside a lambda. But a new replicated lambda instance will create a new App instance depending on load. Also, Lambda warmer will take care of the bootstrapping and caching App Instance in memory. |
|
I try reopen this discussion, yeah its not best practice to have application in lambda, BUT, its possible and devs using this options daily. Great solution for MVP for example, you want check if product is valuable and not working on microservices and complex architecture from start. So why not API for SPA based on AdonisJs in Lambda.. You can check Laravel Vapor https://vapor.laravel.com, where people using Lambda for PHP with whole monolithic apps on serverless without issues. Its easy solution when people can deploy to serverless in few minutes. |
@thetutlage no need to be offensive, I didn't realize that NodeJS is a long running process, not like PHP. Im not skilled in NodeJS, so i cant help yet. See my above comment #1256 (reply in thread) |
No no. God no. I certainly didn't mean to say it like that. |
|
@tzsk Cool. Maybe it was a misunderstanding on my end. Thanks for the clarification 🙂 |
|
Slightly related topic Maybe.. Just maybe something like this can be done in the serverless world. If anyone is more familiar with it, then making AdonisJS |
|
@RomainLanz you are not correct in the statement "Booting-up the application everytime you hit your FaaS URI is extremely inefficient.". |
|
I solved it this way:
const { Ignitor } = require('@adonisjs/ignitor')
const createRequestResponse = require('aws-lambda-create-request-response')
const http = require('http')
var app;
function handler(event, context, callback) {
context.callbackWaitsForEmptyEventLoop =
process.env.WAIT_FOR_EMPTY_EVENT_LOOP === 'yes'
const { req, res } = createRequestResponse(event, callback)
app(req, res)
}
function bootstrapServer() {
return new Ignitor(require('@adonisjs/fold'))
.appRoot(__dirname)
.fireHttpServer((handler) => {
app = handler;
return http.createServer(handler);
})
.catch(console.error);
}
exports.proxy = (event, context, callback) => {
if (app == undefined) {
bootstrapServer().then(() => handler(event, context, callback))
return;
};
handler(event, context, callback);
}
service: service-users
provider:
name: aws
runtime: nodejs12.x
stage: qualidade
region: sa-east-1
functions:
index:
handler: aws-lambda.proxy
events:
- http:
cors: true
path: '/'
method: any
- http:
cors: true
path: '{proxy+}'
method: any
|
|
@tomhatzer I think I found a relevant repo for this: |
|
@capopovici the package this repos is using is this one here: https://github.com/satheler/s12r |
|
Hmm, trying to run the example but I have 2 errors regarding ioc: |
|
Adonis/Core/Server is now here: https://github.com/adonisjs/http-server/blob/develop/src/Server/index.ts Maybe we can ask @satheler if he can make this compatible with the latest version if we can't get it to work. 🤔 |
|
@tomhatzer finallyyyy, solved the issues, will push in a bit |
|
For anyone looking at this. https://github.com/tomhatzer/adonis-serverless
|
|
Woah this is huge! Awesome work @capopovici !! I'll test this today evening too and report anything that may occur that shouldn't occur. |
|
Great work @capopovici. I myself is not much into serverless, but glad to see community pushing limits 🙂 |
|
Neither am I but was interesting and fun to make the framework work with serverless. |
|
@capopovici i‘ll put a readme up with general information and that this is coming from you. |
|
@capopovici I've added a basic README.md file here: https://github.com/tomhatzer/adonis-serverless/blob/master/README.md Can you please check if this is ok for you or otherwise make the necessary changes? |
|
The NODE runtime is tightly coupled with the framework, If we'd want to use it we will have to maintain a forked version of the framework. surveyzop |
|
I've had success for years running node, express and Feathers.js apps on a single Google Firebase/Cloud Function. They also run on node on my local machine. https://github.com/buzzware/function-split-test
|
|
You can run Node.js in AWS lambda serverless functions. I think it would be possible to write an adapter for Adonis applications to work on lambda and other serverless environments. There could be concerns about cold starts but I think it would be interesting and could be a worthwhile tradeoff for some applications. |
|
Spent some time trying to get this to work and I believe I have a solution at least for Adonis Commands running in a docker image on AWS Lambda, I don't think this would be a good idea for serving HTTP requests, more for running one off functions. I plan on using it with SQS triggering Lambda functions to have it process queued jobs. I made a small repo that has some instructions and files to make it easier for others to give it a try. Hopefully it helps some other people out. |
Uh oh!
There was an error while loading. Please reload this page.
The serverless request context is different than normal node server request, because there's no node server in Serverless.
Which is why, if we could do a Serverless adapter for adonis ts that'll be a treat.
Do you have any plans for this in the future? Or recently.
Would be sick if I could run Afonis on a Lambda 🔥😍
All reactions