Deploying to AWS Lambda with ZR CI/CD

On Friday I got started on a very basic set of tools to deploy code to lambda easily.

At ZipRecruiter we don’t use a whole lot of lambda, in general. We have lots of reliable, easy to use infrastructure for things like logging, secrets, etc, and all of that would have to either be reimplemented or just be totally different in Lambda. Be that as it may, we still have code that is best run in Lambda because it is part of bootstrapping our infrastructure.

Today I started on setting up a little Lambda tool that will ensure our central prometheus and alertmanager are running, since if either of those are down, we are flying blind. The tool is a very basic Go program to do a simple web request and then insert a value into CloudWatch. We then set up a CloudWatch alarm to fire if the value is not inserted often enough.

The next step is to actually deploy the code. Normally we’d use Terraform, since this is something running inside of AWS, but because you need to push a zipfile of code to AWS, I thought a better pattern would be to deploy everything to Lambda except for the actual code, and deploy the code itself with the CICD pipeline. The terraform is really basic, and maybe not even helpful, but just in case, here it is:

resource "aws_lambda_function" "monitoring--meta" {
  function_name    = "monitoring--meta"
  role             = "${data.aws_iam_role.lambda.arn}"
  handler          = "meta"
  runtime          = "go1.x"
  filename         = "./"

  vpc_config {
    subnet_ids         = ["${data.aws_subnet_ids.private.ids}"]
    security_group_ids = ["${}"]

  environment {
    variables = {
      ZR_APP_CONFIG = "./conf.json"
} is a zipfile commited to the repo with a single text file in it. If I could have left that out I would have but AWS (or maybe their SDK) doesn’t let you.

Our CICD runs containers in each environment we request them to run in (for example, in prod, or dev, or whatever) and the containers get their own built configuration. While I could have configured the Lambda functions per env, I hardcoded the Lambda config and instead parameterize the config within the zipfile. Here’s the shell script that’s run in the container:


set -x
set -e

cd /tmp
cp /export/monitoring--meta ./monitoring--meta
cp "$ZR_APP_CONFIG" ./config.json

zip ./monitoring--meta ./config.json


With the above we push the meta monitoring binary along with it’s config in a zipfile to lambda. lambda-deploy is a super basic Go command, mostly built to avoid needing a whole Ubuntu image just to push some bytes. Here it is:

package main

import (



type configuration struct {
	Region string

func main() {
	var c configuration
	if err := config.Read(&c); err != nil {
		fmt.Printf("Failed to load config: %s\n", err)

	sess := session.Must(session.NewSession(&aws.Config{
		Region: aws.String(c.Region),
	u := lambda.New(sess)

	b, err := ioutil.ReadFile("")
	if err != nil {
		fmt.Printf("Failed to load zip: %s\n", err)

	_, err = u.UpdateFunctionCode(&lambda.UpdateFunctionCodeInput{
		FunctionName: aws.String("meta--monitoring"),
		Publish:      aws.Bool(true),
		ZipFile:      b,

	if err != nil {
		fmt.Printf("Failed to update function code: %s\n", err)

I think it could have more flag based (or config based) parameterization, but it’s a start and was easy to write. Somewhat incredibly, the first time I ran the lambda uploading code it worked. Now, thanks to this, whenever anyone commits code to master that is in this code or used by this code (ie stuff in common/go) a new build gets triggered and uploaded to lambda. Pretty great.

If you enjoyed this post you might appreciate The SRE Book and likely also The SRE Workbook.

Posted Tue, May 14, 2019

If you're interested in being notified when new posts are published, you can subscribe here; you'll get an email once a week at the most.