How do I use CloudFormation resources in a Lambda function?

Question

I have added a Redis ElastiCache section to my s-resource-cf.json (a CloudFormation template), and selected its hostname as an output.

"Resources": {
    ...snip...
    "Redis": {
        "Type": "AWS::ElastiCache::CacheCluster",
        "Properties": {
            "AutoMinorVersionUpgrade": "true",
            "AZMode": "single-az",
            "CacheNodeType": "cache.t2.micro",
            "Engine": "redis",
            "EngineVersion": "2.8.24",
            "NumCacheNodes": "1",
            "PreferredAvailabilityZone": "eu-west-1a",
            "PreferredMaintenanceWindow": "tue:00:30-tue:01:30",
            "CacheSubnetGroupName": {
                "Ref": "cachesubnetdefault"
            },
            "VpcSecurityGroupIds": [
                {
                    "Fn::GetAtt": [
                        "sgdefault",
                        "GroupId"
                    ]
                }
            ]
        }
    }
},
"Outputs": {
    "IamRoleArnLambda": {
        "Description": "ARN of the lambda IAM role",
        "Value": {
            "Fn::GetAtt": [
                "IamRoleLambda",
                "Arn"
            ]
        }
    },
    "RedisEndpointAddress": {
        "Description": "Redis server host",
        "Value": {
            "Fn::GetAtt": [
                "Redis",
                "Address"
            ]
        }
    }
}

I can get CloudFormation to output the Redis server host when running sls resources deploy, but how can I access that output from within a Lambda function?

There is nothing in this starter project template that refers to that IamRoleArnLambda, which came with the example project. According to the docs, templates are only usable for project configuration, they are not accessible from Lambda functions:

Templates & Variables are for Configuration Only

Templates and variables are used for configuration of the project only. This information is not usable in your lambda functions. To set variables which can be used by your lambda functions, use environment variables.

So, then how do I set an environment variable to the hostname of the ElastiCache server after it has been created?

Answer 1

You can set environment variables in the environment section of a function's s-function.json file. Furthermore, if you want to prevent those variables from being put into version control (for example, if your code will be posted to a public GitHub repo), you can put them in the appropriate files in your _meta/variables directory and then reference those from your s-function.json files. Just make sure you add a _meta line to your .gitignore file.

For example, in my latest project I needed to connect to a Redis Cloud server, but didn't want to commit the connection details to version control. I put variables into my _meta/variables/s-variables-[stage]-[region].json file, like so:

{
  "redisUrl": "...",
  "redisPort": "...",
  "redisPass": "..."
}

…and referenced the connection settings variables in that function's s-function.json file:

"environment": {
  "REDIS_URL": "${redisUrl}",
  "REDIS_PORT": "${redisPort}",
  "REDIS_PASS": "${redisPass}"
}

I then put this redis.js file in my functions/lib directory:

module.exports = () => {
  const redis = require('redis')
  const jsonify = require('redis-jsonify')
  const redisOptions = {
    host: process.env.REDIS_URL,
    port: process.env.REDIS_PORT,
    password: process.env.REDIS_PASS
  }

  return jsonify(redis.createClient(redisOptions))
}

Then, in any function that needed to connect to that Redis database, I imported redis.js:

redis = require('../lib/redis')()

(For more details on my Serverless/Redis setup and some of the challenges I faced in getting it to work, see this question I posted yesterday.)

Answer 2

update

CloudFormation usage has been streamlined somewhat since that comment was posted in the issue tracker. I have submitted a documentation update to http://docs.serverless.com/docs/templates-variables, and posted a shortened version of my configuration in a gist.

It is possible to refer to a CloudFormation output in a s-function.json Lambda configuration file, in order to make those outputs available as environment variables.

s-resource-cf.json output section:

"Outputs": {
    "redisHost": {
        "Description": "Redis host URI",
        "Value": {
            "Fn::GetAtt": [
                "RedisCluster",
                "RedisEndpoint.Address"
            ]
        }
    }
}

s-function.json environment section:

"environment": {
    "REDIS_HOST": "${redisHost}"
},

Usage in a Lambda function:

exports.handler = function(event, context) {
    console.log("Redis host: ", process.env.REDIS_HOST);
};

---

old answer

Looks like a solution was found / implemented in the Serverless issue tracker (link). To quote HyperBrain:

---

CF Output variables

To have your lambda access the CF output variables you have to give it the cloudformation:describeStacks access rights in the lambda IAM role.

The CF.loadVars() promise will add all CF output variables to the process' environment as SERVERLESS_CF_OutVar name. It will add a few ms to the startup time of your lambda.

Change your lambda handler as follows:

// Require Serverless ENV vars
var ServerlessHelpers = require('serverless-helpers-js');
ServerlessHelpers.loadEnv();

// Require Logic
var lib = require('../lib');

// Lambda Handler
module.exports.handler = function(event, context) {
  ServerlessHelpers.CF.loadVars()
  .then(function() {
    lib.respond(event, function(error, response) {
      return context.done(error, response);
    });
  })
};