Can an AWS Lambda function call another

I found a way using the aws-sdk.

var aws = require('aws-sdk');
var lambda = new aws.Lambda({
  region: 'us-west-2' //change to your region
});

lambda.invoke({
  FunctionName: 'name_of_your_lambda_function',
  Payload: JSON.stringify(event, null, 2) // pass params
}, function(error, data) {
  if (error) {
    context.done('error', error);
  }
  if(data.Payload){
   context.succeed(data.Payload)
  }
});

You can find the doc here: http://docs.aws.amazon.com/AWSJavaScriptSDK/latest/AWS/Lambda.html

You should chain your Lambda functions via SNS. This approach provides good performance, latency and scalability for minimal effort.

Your first Lambda publishes messages to your SNS Topic and the second Lambda is subscribed to this topic. As soon as messages arrive in the topic, second Lambda gets executed with the message as it's input parameter.

See Invoking Lambda functions using Amazon SNS notifications.

You can also use this approach to Invoke cross-account Lambda functions via SNS.

here's a sample code for python,

from boto3 import client as boto3_client
from datetime import datetime
import json

lambda_client = boto3_client('lambda')

def lambda_handler(event, context):
    msg = {"key":"new_invocation", "at": datetime.now()}
    invoke_response = lambda_client.invoke(FunctionName="another_lambda_",
                                           InvocationType='Event',
                                           Payload=json.dumps(msg))
    print(invoke_response)

Btw, you would need to add a policy like this to your lambda role as well

   {
        "Sid": "Stmt1234567890",
        "Effect": "Allow",
        "Action": [
            "lambda:InvokeFunction"
        ],
        "Resource": "*"
    }

Since this question was asked, Amazon has released Step Functions (https://aws.amazon.com/step-functions/).

One of the core principles behind AWS Lambda is that you can focus more on business logic and less on the application logic that ties it all together. Step functions allows you to orchestrate complex interactions between functions without having to write the code to do it.

This solution is done using boto3 and Python:

import boto3
import json

lambda_client = boto3.client('lambda', region_name='eu-west-1')

def lambda_handler(event, context):
    lambda_client.invoke(FunctionName = 'function_name', InvocationType = 'RequestResponse', Payload = json.dumps(event))
    
    return True

I was looking at cutting out SNS until I saw this in the Lambda client docs (Java version):

Client for accessing AWS Lambda. All service calls made using this client are blocking, and will not return until the service call completes.

So SNS has an obvious advantage: it's asynchronous. Your lambda won't wait for the subsequent lambda to complete.

Amazon has introduced steps functions in AWS lambda in 2016. I think, now it's more convenient to use steps function as it's really easy to use them. You can build a state machine with two lambda functions as:

to produces a quote

turns a quote into an order

You can easily do that as below:

Here you can have first state for produces a quote and another to turns into order

{
  Comment: "Produce a quote and turns into an order",
  StartAt: "ProduceQuote",
  States: {
    ProduceQuote: {
      "Type": Task,
      "Resource": "arn:aws:lambda:us-east-1:123456789012:function:ProduceQuote",
      "next": TurnsToOrder
    }
    TurnsToOrder: {
      Type: Task,
      Resource: "arn:aws:lambda:us-east-1:123456789012:function:ProduceQuote",
      end: true
    }
  }
}

Steps functions makes it really easy to write multiple lambda functions and run in sequence or in parallel. You can get more information about lambda steps functions here: Steps Functions

In java, we can do as follows :

AWSLambdaAsync awsLambdaAsync = AWSLambdaAsyncClientBuilder.standard().withRegion("us-east-1").build();

InvokeRequest invokeRequest = new InvokeRequest();
invokeRequest.withFunctionName("youLambdaFunctionNameToCall").withPayload(payload);

InvokeResult invokeResult = awsLambdaAsync.invoke(invokeRequest); 

Here, payload is your stringified java object which needs to be passed as Json object to another lambda in case you need to pass some information from calling lambda to called lambda.

There are a lot of answers, but none is stressing that calling another lambda function is not recommended solution for synchronous calls and the one that you should be using is really Step Functions

Reasons why it is not recommended solution:

you are paying for both functions when they are waiting for each other

your code is responsible for retrials

You can also use it for quite complex logic, such as parallel steps and catch failures. Every execution is also being logged out which makes debugging much simpler.

I was working with the answer provided by blueskin but I could not read the Payload response because the InvocationType='Event' is async, so I changed as InvocationType='RequestResponse' and now all works good.

You might be able to make use of the Async.js Waterfall feature - see the bottom part of the big code chunk in Step 3 of this document for an example:

https://aws.amazon.com/blogs/compute/better-together-amazon-ecs-and-aws-lambda/

Others pointed out to use SQS and Step Functions. But both these solutions add additional cost. Step Function state transitions are supposedly very expensive.

AWS lambda offers some retry logic. Where it tries something for 3 times. I am not sure if that is still valid when you trigger it use the API.

Here is the python example of calling another lambda function and gets its response. There is two invocation type 'RequestResponse' and 'Event'. Use 'RequestResponse' if you want to get the response of lambda function and use 'Event' to invoke lambda function asynchronously. So both ways asynchronous and synchronous are available.

    lambda_response = lambda_client.invoke(
                FunctionName = lambda_name,
                InvocationType = 'RequestResponse',
                Payload = json.dumps(input)
                )
    resp_str = lambda_response['Payload'].read()
    response = json.loads(resp_str)

You can invoke lambda function directly (at least via Java) by using AWSLambdaClient as described in the AWS' blog post.

I'm having the same problem but the Lambda function that I implement will insert an entry in DynamoDB, so my solution uses DynamoDB Triggers.

I make the DB invoke a Lambda function for every insert/update in the table, so this separates the implementation of two Lambda functions.

Documentation is here: http://docs.aws.amazon.com/amazondynamodb/latest/developerguide/Streams.Lambda.html

Here is a guided walkthrough: https://aws.amazon.com/blogs/aws/dynamodb-update-triggers-streams-lambda-cross-region-replication-app/

Kind of a roundabout solution but I just call the API endpoint for my lambda functions when I need to chain them. This allows you to decide while coding if you want them to be asynchronous or not.

In case you don't want to setup a POST request you can just setup a simple GET request with a couple, or none at all, query string parameters for easy event passing.

-- Edit --

See: https://docs.aws.amazon.com/apigateway/api-reference/making-http-requests/

and: http://docs.aws.amazon.com/lambda/latest/dg/with-on-demand-https-example.html

You can trigger Lambda functions directly from other Lambda functions directly in an asynchronous manner.

https://docs.aws.amazon.com/lambda/latest/dg/invocation-async.html#invocation-async-destinations

You can set AWS_REGION environment.

assert(process.env.AWS_REGION, 'Missing AWS_REGION env (eg. ap-northeast-1)');
const aws = require('aws-sdk');
const lambda = new aws.Lambda();