Getting Started With Hugo and Lambci

8 minute read Published:


Hugo with LambCI is a combination with a lot of potential: a blog that can be updated from any machine, just by pushing markdown to GitHub. This is entirely achievable with a simple AWS-based toolchain, and no servers required!

This project is fundamentally based on two pieces of software: LambCI and Hugo

LambCI is the work of Michael Hart, who did a great write-up on what it is capable of here

Hugo is an extremely fast static site generator written in (you guessed it) Go, and which is easy to install as a binary executable on most any system. These properties will come in handy later!

Getting started

First thing’s first! You’re gonna need a project dir on your local dev rig, let’s call it: project/
You also need a GitHub repo for project, along with an AWS account.

Setting Up Lambci

Follow the instructions here to get a GitHub token and launch the CloudFormation Stack in AWS

The default configuration utilizes Node. The interpreter will execute the script designated as the handler at the root of your repo.

Configure Lambda

I used the following code in lambci.json to set a shell script to executable, and run it:

  "cmd": "chmod +x && ./"
Note that whatever handler you choose should be located in your repo base dir. In this case, my shell script is in the repo base as well.

I chose to do the heavy lifting in this shell script. If I do this again, I’ll probably select Python over Node (Python is very well suited to these OS/shell automation tasks, and I’m more comfortable with it), and do the entire build process in the Python handler script.

Getting to this point is not that bad. The LambCI CloudFormation template does all the heavy lifting for you, and now you have a build script that will run on the new state of the repo, every time you push to it. With no server, how cool is that?

Wait a Minute… what was Michael saying about ‘Current Limitations’?

Yeah, review those here

The upshot here is that if you really need to do anything serious with LambCI, it’s recommended to use a container service.

In my experience so far this will give you several HUGE advantages, in exchange for more tooling overhead:
1. sudo access (lacking this made installing software in the Lambda environment extremely challenging)
2. unlimited runtime (handy if you have really complex tests, for Hugo this is not an issue)
3. run any AMI you want, useful so you can match to your production environment

There’s more of course, but that’s the obvious stuff. We’ll be OK with basic lambda for now, but if I were to do this again, I’d use ECS for the flexibility and (hopefully) relative ease of use.

With EKS coming out in 2018, I think using an Elastic Container Service is going to get a lot more attractive.

Do You Want To Build a Website

Alright settle down Anna, we’re going to start building your site in project/yourSite. Get started by learning how to use Hugo with the quick start guide

Your goal is to get a working site viewable at localhost:1313 by executing the command hugo server -d in project/yourSite.

Once you’ve gotten that far, running hugo in project/yourSite wil produce project/yourSite/public, which contains your sleek Hugo artifacts, a static web page ready to be served!

Setting up LambCI to use Hugo

We’re going to need to bundle Hugo with the LambCI package so that we can build our Hugo page within the Lambda. This is slightly easier said than done, due to the aforementioned limitations with AWS Lambda.

Building the package:
1. Create a dir lambciPackage
2. clone the lambci repo in ‘lambciPackage’:
3. in lambciPackage/lambci run npm install to get all lambci dependencies
4. grab the latest Hugo_X.XX.X_Linux-64bit.tar.gz from here and put it in lambciPackage/lambci
5. zip uplambciPackage/lambci so that package.json is at the zip root
6. upload the zip package to the lambda function (see the above image of the lambci-build lambda configuration to see where this should be uploaded)

You can now use this prepackaged Hugo tar inside your Lambda Environment

Doing Something Useful With LambCI

This is where things start getting tricky. So, you’ve got at least this in your project root:
project/package.json (or whatever handler you chose to use)
project/ (if you relied on a shell script like me)

So where do these files end up? And that package we uploaded earlier, where the hell is that? And how do we install anything without Sudo??? Lets break it down. When the Lambda environment is started, these directories are the ones you’re gonna care about:
/tmp/lambci/build/**your repo path here** this is where your repo code gets pulled to, and is the initial working dir in the lambda env
/var/task/ is where the package zip file is loaded. Your Hugo tar.gz will be here
/tmp is the wild west, this is where you can install things freely

First thing’s first: we need to move Hugo_X.XX.X_Linux-64bit.tar.gz from the zip package at /var/task to somewhere we can extract it, so let’s copy it to /tmp

Next, untar it with the command tar -xzf /tmp/hugo/$HUGO_RELEASE.tar.gz

Note: I track the Hugo version (Hugo_X.XX.X) in a var in the shell script. This is a good way to make your life less painful if you’re going to refer to it more than once, or ever change it.

Verify Hugo installed correctly by running ./hugo version, and if this spits out a sensible output, you’re ready to build your page! Otherwise, go fish.

Now let’s actually build your site. Use this command:
./hugo -s yourSite
This will put the Hugo output in yourSite/public, where it would be viewable if it only it were on a webserver, and not trapped in this Lambda…

Escape From Lambda: Get To S3!

There are a couple ways to get this going. We need to mirror the directory tree in yourSite/public (in the lambda) into an S3 bucket of your choice, where the world can peruse your ramblings. Furthermore, the contents of that bucket should be set to public-read.

You have some options here: - work out how to install aws-cli in the lambda environment - use the preinstalled AWS SDK for Node, Python, or your lang of choice - install some other wacky S3 interface module into lambda and use that


using aws-cli or a comparable tool is appealing because the command to syncronize your work is dead simple:

aws s3 cp yourSite/public s3://hugosite-artifacts --recursive
there are some pros and cons to this approach. The pro is that it’s really easy to use, and sometimes easy to install. This command worked with certain Lambda interpreter settings:
pip install awscli --upgrade --user
And, there are cons: pip refused to install aws-cli when I tried to build with Node6.10. On top of that, aws-cli requires several dependencies, so it can be a heavy thing to install on the environment unless you will use it really extensively.

I had a devil of a time packaging aws-cli into my Lambda env, it was not as (relatively) painless as Hugo. It’s not clear why this is, but it caused me to use the built in SDK instead.


In this case I will cover the Node SDK. The pro here is that the SDK is built-in to Lambda, so no additional dependencies need to be installed. The con is that copying a directory tree to S3 gets a little more involved than the one-liner used with aws-cli.

I started from the work on this stackoverflow page and made some adjustments to the file params: - set the ACL to public-read (share it with the world) - set the ContentType to the extension of the file (index.html was not opening correctly before I did this)

I went with the recursive solution, so far it has been reliable

Using Some Other Module

Remember way back when we packaging Hugo into your Lambda package? Well right after this step:

3. in `lambciPackage/lambci` run `npm install` to get all lambci dependencies  

We can install a node_module which will interface more nicely with S3!

See You On The Internet

If you will host directly from S3, change your bucket properties to enable static website hosting.

Default settings are fine! Once enabled, an endpoint will be provided from which your static site will be served!

Next: The World

S3 is hosted from a single datacenter, so if folks in Turkmenistan want to read about your coffee roasting habits, it’s helpful to host your site with a more robust CDN (Content Delivery Network).

What comes to mind is AWS CloudFront which caches your page around the world, so traffic to your page is routed to the nearest node (of which there are many), rather than just piling onto that one server in Oregon.

Am I an AWS shill?

Well, if the boot fits.

Credits Roll

Big thanks to Tyler Bacon for helping me troubleshoot AWS IAM and S3 settings. Find him at