AWS IoT Retrieving SQS Messages from a Queue using an IoT Rule and a Lambda function

AWS IoT Rules have predefined rules for sending a message to an SQS Queue, but for some reason not to retrieve a message from a queue using an IoT rule (or does it? if you know how, leave me a comment). You can easily retrieve a message using a Lambda function using the AWS SDK apis though, and you can call a Lambda function using an IoT Rule, so let’s set that up.

To create an IoT Rule to call the Lambda function, which we’ll trigger with incoming MQTT messages on a topic called topic/checkForMessage:

Next, select an action from the long list of available actions, chose ‘Invoke a Lambda function’:

Select the Lambda function we want to call, in this case it’s one created earlier (it has to exist to show in the list, if not press ‘Create a new Resource’ to create one):

On the next summary screen press ‘Create Rule’ and you’re done:

To allow the IoT Rule to call the function, we need to grant the lambda:invokeFunction rule.

Per the docs, we can use the AWS CLI to add the permission like this:

aws lambda add-permission 
  --function-name "function_name" 
  --region "region" 
  --principal iot.amazonaws.com 
  --source-arn arn:aws:iot:us-east-2:account_id:rule/rule_name 
  --source-account "account_id" 
  --statement-id "unique_id" 
  --action "lambda:InvokeFunction"

To apply this to our function and rule, replace:

“function_name” : “LightsOnReceiveMessageFromQueue”

“region”: “us-east-1”

source-arn arn: aarn:aws:iot:full-arn-for-the-rule – see below

account_id: your-aws-account-id

rule_name: RetrieveSQSMessageFromQueue

“account_id”: “your-account-id”

“unique_id”: unique-id-for-this-permission

I’m not sure the AWS Console for IoT shows the ARN for IoT Rules anywhere in it’s pages, but you can easily list it with the AWS CLI, using:

$ aws iot list-topic-rules
{
    "rules": [
        {
            "ruleArn": "arn:aws:iot:us-east-1:your-account-id:rule/RetrieveSQSMessageFromQueue",
            "ruleName": "RetrieveSQSMessageFromQueue",
            "topicPattern": "topic/checkForMessage",
            "createdAt": 1511115896.0,
            "ruleDisabled": false
        }
    ]
}

Ok, plugging in my values into the aws cli statement I have a permission added.

This is it for the IoT Rule. To summarize, this allows us to:

  • respond to incoming messages from an AWS IoT device publishing a message to an MQTT topic called topic/checkForMessages
  • when a message arrives from the device on the topic, it triggers the IoT Rule we just created
  • the rule invokes an AWS Lambda to interact with an AWS SQS Queue to pull a message from a queue.

I’ll share more details on the implementation of the Lambda to interact with the SQS queue and the implementation of the node.js app on a Raspberry Pi in upcoming posts. You’re probably wondering what this is that I’m building? Check back for my followup posts to find out!

This is the second post in a series on AWS and IoT, the first is here:

Installing AWS CLI on MacOS 10.13

The AWS instructions to install the AWS CLI using Python and pip work on MacOS 10.13 (High Sierra) up to the point of adding the Python install location to your path – I found that on 10.13, following the steps didn’t result in the aws command being found.

At the step to addto your path:

  • running ‘which python’ showed:
$ which python
/usr/bin/python

but, ls -la did not show that this was a symbolic link in my install per the docs, so this location is also not the same location where the pip installed aws command is.

This post has an answer that suggests the issue is because AWS CLI instructions tell you to do:

pip3 install awscli --upgrade --user

but the –user option specifies a user location. To find where the pip install is installing to, do:

python3 -m site --user-base

This told me:

/Users/kev/Library/Python/3.6

So taking a looking in the bin dir in this location, this is where the aws cli ended up. Adding this path to my PATH in my ~/.bash_profile and now aws command works as expected.

Publishing a message from a webapp to an AWS SQS Queue via AWS Lamba

I’m building a simple project that needs to receive requests from a simple webpage and process them over time sequentially (more on this later!). Using an AWS SQS queue seems like a good fit for what I’m looking for. Without creating something heavyweight like exposing a REST endpoint running in an EC2 instance, this also seemed like a good opportunity to look into integrating calls from a webpage to a AWS Lambda function. This gives the benefit of not needing to pay for an EC2 instance when it’s up but idle.

To get started I created an AWS SQS queue using the AWS Console (the name of the queue might give away what I’m working on 🙂

I then created a Lambda function to post a message to the queue, using the script from this gist here:

Testing the Lambda from the AWS Console I get this error:

2017-11-12T17:07:19.969Z error: Fail Send Message: AccessDenied: Access to the resource https://sqs.us-east-1.amazonaws.com/ is denied.
2017-11-12T17:07:20.007Z {"errorMessage":"error"}

Per post here, we need to update the default policy we added during creation of the Lambda to include permission to post messages to the queue. The missing permission is to allow sqs:SendMessage and sqs:GetQueueUrl on your SQS Queue resource (insert your ARN for your queue in the Resource name):

{
      "Action": [
        "sqs:SendMessage",
        "sqs:GetQueueUrl"
      ],
      "Effect": "Allow",
      "Resource": "arn:aws:sqs:us-east-1:SOME_ID_HERE:test-messages"
    }

Using the Saved Test Event, now we’re looking good!

2017-11-12T17:32:03.906Z	messageId: f04a...
END RequestId: 658a...
REPORT RequestId: 658a... Duration: 574.09 ms
Billed Duration: 600 ms
Memory Size: 128 MB
Max Memory Used: 42 MB

Let’s take a look in our queue from the SQS Management Console and see if our payload is there:

Now we’ve got our Lambda to post a message into our queue, how can we can call it from a webpage using some Javascript? Looking in the AWS docs there’s an example here. This page also walks through creating configuring  the AWS SDK api to use a Cognito identity pool for unauthorized access to call the Lambda. Step by step on how to create Cognito pools via the AWS Console are in the docs here. It seems there’s a gap in the docs though as it doesn’t explicitly state how to to create a Cognito pool for unauthorized access.

Just out of curiousity, if you attempt to call your Lambda function without any authentication, you get an error that looks like this:

Error: Missing credentials in config
 at credError (bundle.js:10392)
 at Config.getCredentials (bundle.js:10433)
 at Request.VALIDATE_CREDENTIALS (bundle.js:11562)

Ok, so back to creating the Cognito Pool. From the AWS Console, select Cognito. The option you need to select is ‘Manage Federated Identities’ which is where the option is for creating a pool for authenticated access:

Check the box: ‘Enable access to unauthenticated identities’:

Now we’re back to the AWS SDK for JavaScript and can plug in in our Cognito pool id into this config:

AWS.config.update({region: 'REGION'}); AWS.config.credentials = new AWS.CognitoIdentityCredentials({IdentityPoolId: 'IdentityPool'});

My JavaScript to call the Lambda function so far looks like this:

var AWS = require('aws-sdk');

//init AWS credentials with unauthenticated Cognito Identity pool
AWS.config.update({region: 'us-east-1'});
AWS.config.credentials = new AWS.CognitoIdentityCredentials({IdentityPoolId: 'pool-id-here'});

var lambda = new AWS.Lambda();
// create payload for invoking Lambda function
var lambdaCallParams = {
    FunctionName : 'LightsOnMessageToQueue',
    InvocationType : 'RequestResponse',
    LogType : 'None'
};

function callLambda(){
    var result;
    lambda.invoke(lambdaCallParams, function(error, data) {
        if (error) {
            console.log(error);
        } else {
            result = JSON.parse(data.Payload);
            console.log(result);
        }
    });
}

module.exports = {
    callLambda: callLambda
}

Calling the JavaScript now, I get a different error:

assumed-role/Cognito_PostToQueueUnauthRole/CognitoIdentityCredentials
is not authorized to perform: lambda:InvokeFunction 
on resource: arn:aws:lambda:us-east-1:xxx:function:LightsOnMessageToQueue"}

The error is telling us that the permission ‘lambda:InvokeFunction’ is missing for the role Cognito_PostToQueueUnauthRole, so let’s go back and edit and add it. The role was created when we stepped through the Cognito setup steps, but to edit it we need to go to the IAM section on the AWS Console. Searching for Lambda related policies to include in this role, it looks like this is what we’re looking for:

We don’t want to grant InvokeFuntion on all (*) resources though, we can use the JSON for this policy to add a new ‘inline policy’ to our role, and then edit it to specify the ARN for our function.

Back to the JavaScript app, we can now see the SDK making several XHR requests to AWS, including a POST to /functions/LightsOnMessageToQueue/invocations returning with a 200.

Checking the AWS Console, we’re now successfully making calls to our Lambda function, and messages are being posted to the queue:

To host my simple webpage, since it’s static content this can easily be served from AWS S3. I created a new Bucket, granted public read access, and enabled the ‘static website hosting’ website option:

To package the app for deployment, AWS have a sample webpack.config.js here. I did an ‘npm run build’ and then uploaded the index.html and bundle.js to my bucket.

So far this is one part of a project, I’ll post another update when I’ve made some progress on the next part.

 

Installing Docker in an AWS EC2 instance

AWS offers their own EC2 Container Service (ECS) which simplifies deploying Docker containers to EC2 instances (and clusters of instances) and management of your containers. If you want to do-it-yourself though, you can easily install docker yourself in your own instance.

For example, in an Ubuntu EC2 instance,

sudo apt-get install docker.io

Start the docker service with:

sudo service docker start

If you want to manage you own Docker install on EC2, AWS have a guide walking you what you need to know – for further details see here: http://docs.aws.amazon.com/AmazonECS/latest/developerguide/docker-basics.html

(Latest Ubuntu apt packages are docker-ce and docker-ee – see the Docker docs here for more info)