This answer assumes you have already enabled https in the load balancer security group, added the SSL certificate to the load balancer, have both ports 80 and 443 being forwarded by the load balancer, and pointed your domain name at the Elastic Beanstalk environment with Route 53 (or equivalent DNS service).
Option 1: Do the redirect with Apache
This is only possible if you are on an Elastic Beanstalk environment that uses Apache. It may not work for a docker-based deployment.
Amazon Linux 2
Most AWS Linux version 2 based platforms have the option to pick Apache as your proxy host. This can be done by going to "Configuration" > "Software" >
"Container Options" and setting "Proxy Server" to "Apache".
Having done that, add a configuration file named .platform/httpd/conf.d/ssl_rewrite.conf to your codebase (relevant AWS docs) with the following contents:
RewriteEngine On
<If "-n '%{HTTP:X-Forwarded-Proto}' && %{HTTP:X-Forwarded-Proto} != 'https'">
RewriteRule (.*) https://%{HTTP_HOST}%{REQUEST_URI} [R,L]
</If>
Amazon Linux 1
All you need to do is add the following to one of your .config files in the .ebextensions directory of your project:
files:
"/etc/httpd/conf.d/ssl_rewrite.conf":
mode: "000644"
owner: root
group: root
content: |
RewriteEngine On
<If "-n '%{HTTP:X-Forwarded-Proto}' && %{HTTP:X-Forwarded-Proto} != 'https'">
RewriteRule (.*) https://%{HTTP_HOST}%{REQUEST_URI} [R,L]
</If>
Explanation
This is moderately straight forward outside of Elastic Beanstalk. One usually adds an Apache rewrite rule like the following:
RewriteCond %{HTTPS} off
RewriteRule (.*) https://%{HTTP_HOST}%{REQUEST_URI}
Or, if behind a load balancer, like we are in this case:
RewriteCond %{HTTP:X-Forwarded-Proto} !https
RewriteRule (.*) https://%{HTTP_HOST}%{REQUEST_URI} [R,L]
However, these configurations only work within a <VirtualHost> block. Changing the RewriteCond to an <If> block allows it to work properly outside of a <VirtualHost> block, allowing us to put in in a standalone Apache config file. Note that standard Apache setup on CentOS (including the setup on ElasticBeanstalk) inculdes all files matching /etc/httpd/conf.d/*.conf, which matches the file path where we are storing this file.
The -n '%{HTTP:X-Forwarded-Proto}' part of the condition prevents it from redirecting if you are not behind a load balancer, allowing you to have shared configuration between a production evironment with a load balancer and https, and a staging environment that is single instance and does not have https. This is not necessary if you are using load balancers and https on all of your environments, but it doesn't hurt to have it.
Option 2: Do the redirect with the ALB
This is only possible if you are using an Application Load Balancer. Amazon has instructions for how to do that here: https://docs.aws.amazon.com/elasticbeanstalk/latest/dg/configuring-https-httpredirect.html
All you need to do is add the following to one of your .config files in the .ebextensions directory of your project to replace the http listener with a redirect:
Resources:
AWSEBV2LoadBalancerListener:
Type: AWS::ElasticLoadBalancingV2::Listener
Properties:
LoadBalancerArn:
Ref: AWSEBV2LoadBalancer
Port: 80
Protocol: HTTP
DefaultActions:
- Type: redirect
RedirectConfig:
Host: "#{host}"
Path: "/#{path}"
Port: "443"
Protocol: "HTTPS"
Query: "#{query}"
StatusCode: "HTTP_301"
Bad solutions I have seen
I have seen a lot of bad solutions to this problem, and it is worth going through them to understand why this solution is necessary.
Use Cloudfront: Some people suggest using non-cached Cloudfront setup in front of Elastic Beanstalk to do the HTTP to HTTPS redirect. This adds a whole new service (thus adding complexity) that isn't exactly appropriate (Cloudfront is a CDN; it's not the right tool for forcing HTTPS on inherantly dynamic content). Apache config is the normal solution to this problem and Elastic Beanstalk uses Apache, so that's the way we should go.
SSH into the server and...: This is completely antithetical to the point of Elastic Beanstalk and has so many problems. Any new instances created by autoscaling won't have the modified configuration. Any cloned environments won't have the configuration. Any number of a reasonable set of environment changes will wipe out the configuration. This is just such a bad idea.
Overwrite the Apache config with a new file: This is getting into the right realm of solution but leaves you with a maintenance nightmare if Elastic Beanstalk changes aspects of the server setup (which they very well may do). Also see the problems in the next item.
Dynamically edit the Apache config file to add a few lines: This is a decent idea. The problems with this is that it won't work if Elastic Beanstalk ever changes the name of their default Apache config file, and that this file can get overwritten when you least expect: https://forums.aws.amazon.com/thread.jspa?threadID=163369
