Why is this HTTP request not working on AWS Lambda

问题:

I'm getting started with AWS Lambda and I'm trying to request an external service from my handler function. According to this answer, HTTP requests should work just fine, and I haven't found any documentation that says otherwise. (In fact, people have posted code that use the Twilio API to send SMS.)

My handler code is:

var http = require('http');

exports.handler = function(event, context) {
  console.log('start request to ' + event.url)
  http.get(event.url, function(res) {
    console.log("Got response: " + res.statusCode);
  }).on('error', function(e) {
    console.log("Got error: " + e.message);
  });

  console.log('end request to ' + event.url)
  context.done(null);
}

and I see the following 4 lines in my CloudWatch logs:

2015-02-11 07:38:06 UTC START RequestId: eb19c89d-b1c0-11e4-bceb-d310b88d37e2
2015-02-11 07:38:06 UTC eb19c89d-b1c0-11e4-bceb-d310b88d37e2 start request to http://www.google.com
2015-02-11 07:38:06 UTC eb19c89d-b1c0-11e4-bceb-d310b88d37e2 end request to http://www.google.com
2015-02-11 07:38:06 UTC END RequestId: eb19c89d-b1c0-11e4-bceb-d310b88d37e2

I'd expect another line in there:

2015-02-11 07:38:06 UTC eb19c89d-b1c0-11e4-bceb-d310b88d37e2 Got response: 302

but that's missing. If I'm using the essential part without the handler wrapper in node on my local machine, the code works as expected.

The inputfile.txt I'm using is for the invoke-async call is this:

{
   "url":"http://www.google.com"
}

It seems like the part of the handler code that does the request is skipped entirely. I started out with the request lib and fell back to using plain http to create a minimal example. I've also tried to request a URL of a service I control to check the logs and there's no requests coming in.

I'm totally stumped. Is there any reason Node and/or AWS Lambda would not execute the HTTP request?

回答1:

Of course, I was misunderstanding the problem. As AWS themselves put it:

For those encountering nodejs for the first time in Lambda, a common error is forgetting that callbacks execute asynchronously and calling context.done() in the original handler when you really meant to wait for another callback (such as an S3.PUT operation) to complete, forcing the function to terminate with its work incomplete.

I was calling context.done way before any callbacks for the request fired, causing the termination of my function ahead of time.

The working code is this:

var http = require('http');

exports.handler = function(event, context) {
  console.log('start request to ' + event.url)
  http.get(event.url, function(res) {
    console.log("Got response: " + res.statusCode);
    context.succeed();
  }).on('error', function(e) {
    console.log("Got error: " + e.message);
    context.done(null, 'FAILURE');
  });

  console.log('end request to ' + event.url);
}

Update: starting 2017 AWS has deprecated the old Nodejs 0.10 and only the newer 4.3 run-time is now available (old functions should be updated). This runtime introduced some changes to the handler function. The new handler has now 3 parameters.

function(event, context, callback)

Although you will still find the succeed, done and fail on the context parameter, AWS suggest to use the callback function instead or null is returned by default.

callback(new Error('failure')) // to return error
callback(null, 'success msg') // to return ok

Complete documentation can be found at http://docs.aws.amazon.com/lambda/latest/dg/nodejs-prog-model-handler.html

回答2:

Yeah, awendt answer is perfect. I'll just show my working code... I had the context.succeed('Blah'); line right after the reqPost.end(); line. Moving it to where I show below solved everything.

console.log('GW1');

var https = require('https');

exports.handler = function(event, context) {

    var body='';
    var jsonObject = JSON.stringify(event);

    // the post options
    var optionspost = {
        host: 'the_host',
        path: '/the_path',
        method: 'POST',
        headers: {
            'Content-Type': 'application/json',
        }
    };

    var reqPost = https.request(optionspost, function(res) {
        console.log("statusCode: ", res.statusCode);
        res.on('data', function (chunk) {
            body += chunk;
        });
        context.succeed('Blah');
    });

    reqPost.write(jsonObject);
    reqPost.end();
};

回答3:

I had the very same problem and then I realized that programming in NodeJS is actually different than Python or Java as its based on JavaScript. I'll try to use simple concepts as there may be a few new folks that would be interested or may come to this question.

Let's look at the following code :

var http = require('http'); // (1)
exports.handler = function(event, context) {
  console.log('start request to ' + event.url)
  http.get(event.url,  // (2)
  function(res) {  //(3)
    console.log("Got response: " + res.statusCode);
    context.succeed();
  }).on('error', function(e) {
    console.log("Got error: " + e.message);
    context.done(null, 'FAILURE');
  });

  console.log('end request to ' + event.url); //(4)
}

Whenever you make a call to a method in http package (1) , it is created as event and this event gets it separate event. The 'get' function (2) is actually the starting point of this separate event.

Now, the function at (3) will be executing in a separate event, and your code will continue it executing path and will straight jump to (4) and finish it off, because there is nothing more to do.

But the event fired at (2) is still executing somewhere and it will take its own sweet time to finish. Pretty bizarre, right ?. Well, No it is not. This is how NodeJS works and its pretty important you wrap your head around this concept. This is the place where JavaScript Promises come to help.

You can read more about JavaScript Promises here. In a nutshell, you would need a JavaScript Promise to keep the execution of code inline and will not spawn new / extra threads.

Most of the common NodeJS packages have a Promised version of their API available, but there are other approaches like BlueBirdJS that address the similar problem.

The code that you had written above can be loosely re-written as follows.

'use strict';
console.log('Loading function');
var rp = require('request-promise');
exports.handler = (event, context, callback) => {    

    var options = {
    uri: 'https://httpbin.org/ip',
    method: 'POST',
    body: {

    },
    json: true 
};


    rp(options).then(function (parsedBody) {
            console.log(parsedBody);
        })
        .catch(function (err) {
            // POST failed... 
            console.log(err);
        });

    context.done(null);
};

Please note that the above code will not work directly if you will import it in AWS Lambda. For Lambda, you will need to package the modules with the code base too.

回答4:

I've found lots of posts across the web on the various ways to do the request, but none that actually show how to process the response synchronously on AWS Lambda.

Here's a Node 6.10.3 lambda function that uses an https request, collects and returns the full body of the response, and passes control to an unlisted function processBody with the results. I believe http and https are interchangable in this code.

I'm using the async utility module, which is easier to understand for newbies. You'll need to push that to your AWS Stack to use it (I recommend the serverless framework).

Note that the data comes back in chunks, which are gathered in a global variable, and finally the callback is called when the data has ended.

'use strict';

const async = require('async');
const https = require('https');

module.exports.handler = function (event, context, callback) {

    let body = "";
    let countChunks = 0;

    async.waterfall([
        requestDataFromFeed,
        // processBody,
    ], (err, result) => {
        if (err) {
            console.log(err);
            callback(err);
        }
        else {
            const message = "Success";
            console.log(result.body);
            callback(null, message);
        }
    });

    function requestDataFromFeed(callback) {
        const url = 'https://put-your-feed-here.com';
        console.log(`Sending GET request to ${url}`);
        https.get(url, (response) => {
            console.log('statusCode:', response.statusCode);
            response.on('data', (chunk) => {
                countChunks++;
                body += chunk;
            });
            response.on('end', () => {
                const result = {
                    countChunks: countChunks,
                    body: body
                };
                callback(null, result);
            });
        }).on('error', (err) => {
            console.log(err);
            callback(err);
        });
    }
};

回答5:

Yes, there's in fact many reasons why you can access AWS Lambda like and HTTP Endpoint.

The architecture of AWS Lambda

It's a microservice. Running inside EC2 with Amazon Linux AMI (Version 3.14.26–24.46.amzn1.x86_64) and runs with Node.js. The memory can be beetwen 128mb and 1gb. When the data source triggers the event, the details are passed to a Lambda function as parameter's.

What happen?

AWS Lambda run's inside a container, and the code is directly uploaded to this container with packages or modules. For example, we NEVER can do SSH for the linux machine running your lambda function. The only things that we can monitor are the logs, with CloudWatchLogs and the exception that came from the runtime.

AWS take care of launch and terminate the containers for us, and just run the code. So, even that you use require('http'), it's not going to work, because the place where this code runs, wasn't made for this.