berfr blog

Deploying to S3 with Sourcehut

Posted on February 05, 2020.

In this article, we will go through the process of automatically building and deploying this Hugo website to Amazon S3 using Sourcehut This post assumes you already have a static website that is hosted on S3; there are plenty of guides online concerning this if it is not the case.

IAM policy for S3 bucket access

The first step is to create an IAM user with an IAM policy that can only modify the contents of the specific S3 bucket. This will ensure that if the user credentials are compromised, only the specific bucket is affected and not your complete AWS account. This is important since this user secret key will be saved on Sourcehut which is not immune to security bugs and breaches.

Once the new user is created, note its Access Key ID and Secret Access Key. We will be using these to access AWS from the CI environment.

We will be using the s3 sync command from the AWS CLI tool which requires specific permissions to work. To get it to execute without errors, the following policy is attached directly to the IAM user:

    "Version": "2012-10-17",
    "Statement": [
            "Sid": "VisualEditor0",
            "Effect": "Allow",
            "Action": [
            "Resource": [

Sourcehut secrets

We will now copy the IAM user credentials noted earlier into Sourcehut build secrets so it is available during CI builds. Create a new secret of File type at ~/.aws/credentials with the content in this format:


You can give the file minimum read permissions 400. Save the secret and note the resulting UUID; we will use it in the build manifest file.

Sourcehut build manifest

The next step will be to create a build manifest: the .build.yml file in the root of the repository with the information and instructions to build and deploy the static website. Here are the contents for this specific site:

image: fedora/31
    - awscli
    - hugo
    - 1af67a38-6cf0-4f09-98fc-8776b8987160
    - build: |
        make build
        git describe --exact-match HEAD || complete-build
    - deploy: |
        make deploy

As you can see, the file is pretty simple. Here is a description of some of the components:

The Makefile used is also easy to follow:

.PHONY: build deploy release

	hugo --cleanDestinationDir --minify

deploy: build
	aws --region ca-central-1 --profile s3 sync public/ s3:// --delete

release: build
	git tag -s -a "`date +\"%y-%m-%d-%H-%M\"`" -m "`date`"

By using a Makefile, we can make changes to the commands needed in a single file and use it both locally and in the build manifest. The same goes for specifying a unique AWS profile name such as; we are able to use it in both environments.

With these two files pushed to, the build should automatically start and be visible at To trigger a deployment, simply run make release followed by git push --follow-tags.

You can find the complete source code for this webite at and the builds for it at If you have any questions or comments regarding this article, feel free to email me personally or write to my public inbox.