582 lines (402 sloc) 23.2 KB

Deploy your website to S3

Build Status Gem Version

What s3_website can do for you

  • Create and configure an S3 website for you
  • Upload your static website to AWS S3
  • Jekyll, Nanoc, and Middleman are automatically supported
  • Help you use AWS Cloudfront to distribute your website
  • Improve page speed with HTTP cache control and gzipping
  • Set HTTP redirects for your website
  • (for other features, see the documentation below)


gem install s3_website

s3_website needs both Ruby and Java to run. (S3_website is partly written in Scala, hence the need for Java.)


Here's how you can get started:

  • Create API credentials that have sufficient permissions to S3. More info here.
  • Go to your website directory
  • Run s3_website cfg create. This generates a configuration file called s3_website.yml.
  • Put your AWS credentials and the S3 bucket name into the file
  • Run s3_website cfg apply. This will configure your bucket to function as an S3 website. If the bucket does not exist, the command will create it for you.
  • Run s3_website push to push your website to S3. Congratulations! You are live.
  • At any later time when you would like to synchronise your local website with the S3 website, simply run s3_website push again. (It will calculate the difference, update the changed files, upload the new files and delete the obsolete files.)

Specifying the location of your website

S3_website will automatically discover websites in the _site and public/output directories.

If your website is not in either of those directories, you can point the location of your website in two ways:

  1. Add the line site: path-to-your-website into the s3_website.yml file
  2. Or, use the --site=path-to-your-site command-line argument

If you want to store the s3_website.yml file in a directory other than the project's root you can specify the directory like so: s3_website push --config-dir config.

Using standard AWS credentials

If you omit s3_id from your s3_website.yml, S3_website will fall back to reading from the default AWS SDK locations. For instance, if you've used aws configure to set up credentials in ~/.aws/credentials, S3_website can use these.

Using environment variables

You can use ERB in your s3_website.yml file which incorporates environment variables:

s3_id: <%= ENV['S3_ID'] %>
s3_secret: <%= ENV['S3_SECRET'] %>

(If you are using s3_website on an EC2 instance with IAM roles, you can omit the s3_id and s3_secret keys in the config file.)

S3_website implements support for reading environment variables from a file using the dotenv gem. You can create a .env file in the project's root directory to take advantage of this feature. Please have a look at dotenv's usage guide for syntax information.

Your .env file should containing the following variables:


S3_website will also honor environment variables named AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_SESSION_TOKEN (if using STS) automatically if s3_id is ommitted from s3_website.yml.

Project goals

  • Provide a command-line interface tool for deploying and managing S3 websites
  • Let the user have all the S3 website configurations in a file
  • Minimise or remove the need to use the AWS Console
  • Allow the user to deliver the website via CloudFront
  • Automatically detect the most common static website tools, such as Jekyll, Nanoc, and Middleman.
  • Be simple to use: require only the S3 credentials and the name of the S3 bucket
  • Let the power users benefit from advanced S3 website features such as redirects, Cache-Control headers and gzip support
  • Be as fast as possible. Do in parallel all that can be done in parallel.

s3_website attempts to be a command-line interface tool that is easy to understand and use. For example, s3_website --help should print you all the things it can perform. Please create an issue if you think the tool is incomprehensible or inconsistent.

Additional features

Cache Control

You can use either the setting max_age or cache_controlto enable more effective browser caching of your static assets.


There are two possible ways to use the option: you can specify a single age (in seconds) like so:

max_age: 300

Or you can specify a hash of globs, and all files matching those globs will have the specified age:

  "assets/*": 6000
  "*": 300

After changing the max_age setting, push with the --force option. Force-pushing allows you to update the S3 object metadata of existing files.


The cache_control setting allows you to define an arbitrary string that s3_website will put on all the S3 objects of your website.

Here's an example:

cache_control: public, no-transform, max-age=1200, s-maxage=1200

You can also specify a hash of globs, and all files matching those globs will have the specified cache-control string:

  "assets/*": public, max-age=3600
  "*": no-cache, no-store

After changing the cache_control setting, push with the --force option. Force-pushing allows you to update the S3 object metadata of existing files.

Content type detection

By default, s3_website automatically detects the content type of a file with the help of Apache Tika.

For some file types Tika's auto detection does not work correctly. Should this problem affect you, use the content_type setting to override Tika's decision:

  "*.myextension": application/my-custom-type

Gzip Compression

If you choose, you can use compress certain file types before uploading them to S3. This is a recommended practice for maximizing page speed and minimizing bandwidth usage.

To enable Gzip compression, simply add a gzip option to your s3_website.yml configuration file:

gzip: true

Note that you can additionally specify the file extensions you want to Gzip (.html, .css, .js, .ico, and .txt will be compressed when gzip: true):

  - .html
  - .css
  - .md

Remember that the extensions here are referring to the compiled extensions, not the pre-processed extensions.

After changing the gzip setting, push with the --force option.

s3_website will not gzip a file that is already gzipped. This is useful in the situations where your build tools gzip a file before you invoke s3_website push.

Using non-standard AWS regions

By default, s3_website uses the US Standard Region. You can upload your website to other regions by adding the setting s3_endpoint into the s3_website.yml file.

For example, the following line in s3_website.yml will instruct s3_website to push your site into the Tokyo region:

s3_endpoint: ap-northeast-1

The valid s3_endpoint values consist of the S3 location constraint values.

Ignoring files you want to keep on AWS

Sometimes there are files or directories you want to keep on S3, but not on your local machine. You may define a regular expression to ignore files like so:

ignore_on_server: that_folder_of_stuff_i_dont_keep_locally

You may also specify the values as a list:

  - that_folder_of_stuff_i_dont_keep_locally
  - file_managed_by_somebody_else

If you add the magic word ignore_on_server: _DELETE_NOTHING_ON_THE_S3_BUCKET_, s3_website push will never delete any objects on the bucket.

Excluding files from upload

You can instruct s3_website not to push certain files:

exclude_from_upload: test

The value can be a regex, and you can specify many of them:

  - test
  - (draft|secret)

Reduced Redundancy

You can reduce the cost of hosting your blog on S3 by using Reduced Redundancy Storage:

  • In s3_website.yml, set s3_reduced_redundancy: true
  • All objects uploaded after this change will use the Reduced Redundancy Storage.
  • If you want to change all of the files in the bucket, you can change them through the AWS console, or update the timestamp on the files before running s3_website again

After changing the s3_reduced_redundancy setting, push with the --force option.

How to use Cloudfront to deliver your blog

It is easy to deliver your S3-based web site via Cloudfront, the CDN of Amazon.

Creating a new CloudFront distribution

When you run the command s3_website cfg apply, it will ask you whether you want to deliver your website via CloudFront. If you answer yes, the command will create a CloudFront distribution for you.

If you do not want to receive this prompt, or if you are running the command in a non-interactive session, you can use s3_website cfg apply --headless (and optionally also use --autocreate-cloudfront-dist if desired).

Using your existing CloudFront distribution

If you already have a CloudFront distribution that serves data from your website S3 bucket, just add the following line into the file s3_website.yml:

cloudfront_distribution_id: your-dist-id

Next time you run s3_website push, it will invalidate the items on CloudFront and thus force the CDN system to reload the changes from your website S3 bucket.

Specifying custom settings for your CloudFront distribution

s3_website lets you define custom settings for your CloudFront distribution.

For example, like this you can define your own TTL and CNAME:

    min_ttl: <%= 60 * 60 * 24 %>
    quantity: 1

Once you've saved the configuration into s3_website.yml, you can apply them by running s3_website cfg apply.

Invalidating all CloudFront resources (wildcard invalidation)

The following setting is recommended for most users:

cloudfront_wildcard_invalidation: true

Over time, it can reduce your AWS bill significantly. For more information, see

Invalidating root resources instead of index.htmls

By default, s3_website push calls the CloudFront invalidation API with the file-name-as-it-is. This means that if your file is article/index.html, the push command will call the invalidation API on the resource article/index.html.

You can instruct the push command to invalidate the root resource instead of the index.html resource by adding the following setting into the configuration file:

cloudfront_invalidate_root: true

To recap, this setting instructs s3_website to invalidate the root resource (e.g., article/) instead of the filename'd resource (e.g., article/index.html).

No more index.htmls in your URLs!

Note: If the root resource on your folder displays an error instead of the index file, your source bucket in Cloudfront likely is pointing to the S3 Origin, Update the source to the S3 Website Endpoint, e.g., to fix this.

Configuring redirects on your S3 website

You can set HTTP redirects on your S3 website in three ways.

Exact page match for moving a single page

If a request is received matching a string e.g. /heated-towel-rail/ redirect to a page e.g. /

This kind of redirect is created in the s3_website.yml file under the redirects: section as follows:

  index.php: /
  about.php: /about.html
  heated-towel-rail/index.html: /

Note that the root forward slash is omitted in the requested page path and included in the redirect-to path. Note also that in the heated-towel-rail example this also matches heated-towel-rail/ since S3 appends index.html to request URLs terminated with a slash.

This redirect will be created as a 301 redirect from the first URL to the destination URL on the same server with the same http protocol.

Under the hood s3_website creates a zero-byte index.html page for each path you want redirected with the appropriate x-amz-website-redirect-location value in the metadata for the object. See Amazon S3's x-amz-website-redirect-location documentation for more details.

On terminology: the left value is the redirect source and the right value is the redirect target. For example above, about.php is the redirect source and /about.html the target.

If the s3_key_prefix setting is defined, it will be applied to the redirect target if and only if the redirect target points to a site-local resource and does not start with a slash. E.g., about.php: about.html will be translated into about.php: VALUE-OF-S3_KEY_PREFIX/about.html.

Prefix replacement for moving a folder of pages

Common to content migrations, content pages often move from one subdirectory to another. For example if you're moving all the case studies on your site under /portfolio/work/ to /work/. In this case we use a prefix replacement such that /portfolio/work/walkjogrun/ gets 301 redirected to /work/walkjogrun/.

To do this we add a new rule to the routing_rules: section as follows:

  - condition:
        key_prefix_equals: portfolio/work/
        protocol: https
        host_name: <%= ENV['REDIRECT_DOMAIN_NAME'] %>
        replace_key_prefix_with: work/
        http_redirect_code: 301


  • -condition: indicates the start of a new rule.
  • key_prefix_equals: introduces the path prefix (also without the leading / per the exact page match). Note that this prefix matches anything underneath it so every case study under that path will be handled by the subsequent redirect
  • redirect: indicates the start of the redirect definition
  • protocol: is optional and defaults to http.
  • host_name: is optional but the default is the bucket name not the actual domain name so this also effectively required for our site. In this example we use an environment variable to store the server hostname to support building to different environments. REDIRECT_DOMAIN_NAME can be configured on a CI server as well any CodePipelines responsible for building the site to different environments. If you're running locally you'll need to set
  • replace_key_prefix_with: indicates the substitution to use in place of the matched prefix. This is the only field required by s3_website, so effectively this rule works like a replace rule e.g. replace portfolio/work with /work in the string portfolio/work/walkjogrun
  • http_redirect_code: is optional and defaults to 302 Temporary redirect which is terrible for SEO since your content temporarily vanishes from the Google index until the response changes for the URL. This is almost never what you want. You can use this to temporarily redirect any content you haven't migrated to the new site yet as long as you remove or replace the 302 with a link to a permanent home. This tells Google to forget the old location of the page and use the new content at the new URL. For pages that move you'll see little if any discrepancy in Google traffic.

After adding the configuration, run the command s3_website cfg apply on your command-line interface. This will apply the routing rules on your S3 bucket.

For more information on configuring redirects, see the documentation of the configure-s3-website gem, which comes as a transitive dependency of the s3_website gem. (The command s3_website cfg apply internally calls the configure-s3-website gem.)

Prefix coallescing for deleting pages (or consolidating)

If you 301 redirect lots of content into one new path you're telling Google that the old pages are gone so only the destination page is important moving forward. E.g. if you had 10 services pages and consolidate them into 1 services listing page you'll lose the 10 pages uniquely optimized for different sets of keywords and retain just 1 page with no real keyword focus and hence less SEO value.

For example, we're not porting the entire set of pages under the folder /experience to the new website. Some of these pages still get traffic from either Google or inbound links so we don't want to just show a 404 content not found error. We will 301 redirect them to the most useful replacement page. In the case of /experience we don't have anything better to show than just the home page so that is how the redirect is configured.

Here's how to redirect to indicate a deleted page:

  - condition:
        key_prefix_equals: experience/
        protocol: https
        host_name: <%= ENV['REDIRECT_DOMAIN_NAME'] %>
        **replace_key_with**: /
        http_redirect_code: 301

Note the only difference is that instead of using replace_key_prefix_with we use replace_key_with to effectively say "replace the entire path matching the prefix specfied in the condition with the new path".

On skipping application of redirects

If your website has a lot of redirects, you may find the following setting helpful:

treat_zero_length_objects_as_redirects: true

The setting allows s3_website push to infer whether a redirect exists on the S3 bucket. You will experience faster push performance when this setting is true.

If this setting is enabled and you modify the redirects setting in s3_website.yml, use push --force to apply the modified values.

For backward-compatibility reasons, this setting is false by default.

In this context, the word object refers to object on S3, not file-system file.

Specifying custom concurrency level

By default, s3_website does 3 operations in parallel. An operation can be an HTTP PUT operation against the S3 API, for example.

You can increase the concurrency level by adding the following setting into the s3_website.yml file:

concurrency_level: <integer>

However, because S3 throttles connections, there's an upper limit to the level of parallelism. If you start to see end-of-file errors, decrease the concurrency level. Conversely, if you don't experience any errors, you can increase the concurrency level and thus benefit from faster uploads.

If you experience the "too many open files" error, either increase the amount of maximum open files (on Unix-like systems, see man ulimit) or decrease the concurrency_level setting.

Simulating deployments

You can simulate the s3_website push operation by adding the --dry-run switch. The dry run mode will not apply any modifications on your S3 bucket or CloudFront distribution. It will merely print out what the push operation would actually do if run without the dry switch.

You can use the dry run mode if you are unsure what kind of effects the push operation would cause to your live website.

S3 website in a subdirectory of the bucket

If your S3 website shares the same S3 bucket with other applications, you can push your website into a "subdirectory" on the bucket.

Define the subdirectory like so:

s3_key_prefix: your-subdirectory

Temporary security credentials with Session Token

AWS temporary security credentials (eg: when assuming IAM roles)


session_token: your-token

Migrating from v1 to v2

Please read the release note on version 2. It contains information on backward incompatible changes.

You can find the v1 branch here. It's in maintenance mode. This means that v1 will see only critical bugfix releases.

Example configurations

See more example-configurations

On security

If the source code of your website is publicly available, ensure that the s3_website.yml file is in the list of ignored files. For git users this means that the file .gitignore should mention the s3_website.yml file.

If you use the .dotenv gem, ensure that you do not push the .env file to a public git repository.

Known issues

Please create an issue and send a pull request if you spot any.


s3_website uses Semantic Versioning.

In the spirit of semantic versioning, here is the definition of public API for s3_website: Within a major version, s3_website will not break backwards-compatibility of anything that is mentioned in this README file.


See development.


We (users and developers of s3_website) welcome patches, pull requests and ideas for improvement.

When sending pull requests, please accompany them with tests. Favor BDD style in test descriptions. Use VCR-backed integration tests where possible. For reference, you can look at the existing s3_website tests.

If you are not sure how to test your pull request, you can ask the gem owners to supplement the request with tests. However, by including proper tests, you increase the chances of your pull request being incorporated into future releases.



MIT. See the LICENSE file for more information.


This gem is created by Lauri Lehmijoki. Without the valuable work of Philippe Creux on jekyll-s3, this project would not exist.

See the Contributors.

Community articles


Support via Gittip