This project is forked from the AWS project providing a reference architecture for Drupal.
It is modified from that fork to:
- use PHP 7.4 instead of PHP 7.0
- use the latest Amazon Linux 2 AMIs
- use modern instance types
- support cheaper RDS options for testing
- install Drupal with
composer
- install the dependencies needed by LocalGovDrupal
- install LocalGovDrupal itself via
drush
You can launch this CloudFormation stack, using your account, in the following AWS Regions:
AWS Region Code | Name | Launch |
---|---|---|
us-east-1 | US East (N. Virginia) | ![]() |
us-east-2 | US East (Ohio) | ![]() |
us-west-2 | US West (Oregon) | ![]() |
eu-west-1 | EU (Ireland) | ![]() |
eu-west-2 | UK (London) | ![]() |
ap-southeast-2 | AP (Sydney) | ![]() |
Installation takes a while (composer install
then drush si
) so you'll have to be quite patient - if you're still getting Bad Gateway after half an hour or so it's time to investigate, before then it could still be building, as the install process happens on initialization of the machines.
Because of disk caching the initial cluster can fail to build if there is more than one web server (first one will "win" and second one will end in a bad state because the Apache start command won't get called). There is a random sleep to try and get around this, whereby both servers are asked to wait for a random delay up to 60 seconds. Normally this is sufficient to make sure they don't both try to compose install
at the same time, but if it fails you can get Bad Gateway errors as one of the machines will be left in a bad state. In such cases just change your Autoscaling Group size and the newly generated machines should be fine. Alternatively, edit the Bastion ASG and spin up a bastion machine, SSH to it and from there you can hop on to the broken web server and start Apache with sudo service httpd start
.
Drupal is installed on EFS, which may be slow. This is the recommended approach by AWS, on the basis that CloudFront should be taking up most of the slack anyway. There are known but clunky workarounds for slow EFS.
ElastiCache and the necessary Apache module are installed if you select that option, but the Drupal installation process does not do any memcached set-up, so if you want to make use of ElastiCache you'll need to manually set Drupal up. (AWS wrote docs on that you'll find below.)
From this point onward everything is taken from the original AWS reference stack README.
This reference architecture provides a set of YAML templates for deploying LocalGovDrupal on AWS using Amazon Virtual Private Cloud (Amazon VPC), Amazon Elastic Compute Cloud (Amazon EC2), Auto Scaling, Elastic Load Balancing (Application Load Balancer), Amazon Relational Database Service (Amazon RDS), Amazon ElastiCache, Amazon Elastic File System (Amazon EFS), Amazon CloudFront, Amazon Route 53, Amazon Certificate Manager (Amazon ACM) with AWS CloudFormation.
The repository consists of a set of nested templates which are run in order from the master template. Run the master template to create the entire stack, entering the appropriate parameters. Nested templates can be run individually in order, entering the appropriate input parameters for each stack.
To launch the entire stack and deploy a Drupal site on AWS, click on one of the Launch Stack links above or download the Master template and launch it locally.
AWS Certificate Manager (ACM) is a service that lets you easily provision, manage, and deploy Secure Sockets Layer/Transport Layer Security (SSL/TLS) certificates for use with AWS services. SSL/TLS certificates provisioned through AWS Certificate Manager are free.
If you don't already have an SSL/TLS certificate for your domain name, it is recommended that you request one using ACM. For more information about requesting an SSL/TLS certificate using ACM, please read the AWS Certificate Manager User Guide.
Use ACM to request a certificate or import a certificate into ACM. To use an ACM certificate with CloudFront (optional input parameter), you must request or import the certificate in the US East (N. Virginia) region. To use an ACM certificate with Amazon ELB - Application Load Balancer (optional input parameter), you must request or import the certificate in the region you create the CloudFormation stack. After you validate ownership of the domain names in your certificate, ACM provisions the certificate. Use the ACM certificate Amazon Resource Name (ARN) as the optional Cloudfront and/or Public ALB ACM certificate input parameters of the master template.
AWS Certificate Manager (ACM) is a service that lets you easily provision, manage, and deploy Secure Sockets Layer/Transport Layer Security (SSL/TLS) certificates for use with AWS services. SSL/TLS certificates provisioned through AWS Certificate Manager are free.
There are two output URLs for the master template. The SiteURL will take you to your new Drupal site installation wizard. It will be the site domain name if you provided one, the CloudFront URL if you chose to create a CloudFront distribution, or the domain name of the public application load balancer. The second output URL will take you to the OpCache Status page for each EC2 instance in the AutoScaling group. By refreshing the page you will be able to verify OpCache has been enabled on each instance.
Before starting the installation wizard, verify that the OpCache status page is available for all EC2 instances in the AutoScaling group by clicking on the OpCacheValidationURL link of the master template's Output tab. Refresh this page a few times to verify that the EC2 instance id changes.
To start the installation wizard, click on the SiteURL link of the master template's Output tab.
Select the appropriate language and click Save and continue.
Select your desired installation profile and click Save and continue.
Enter your database configuration. Enter the Database name, Database username, and Database password you entered as parameters of the CloudFormation master template. Select ADVANCED OPTIONS and enter the DatabaseClusterEndpointAddress of the RDS cluster created with this stack. This can be found in the CloudFormation Stack window by selecting the RDS statck that was just created and selecting the Output tab. Click Save and continue.
The installation wizard will install the site.
After the site has been installed, enter the site information and click Save and continue.
OPcache is a byte-code cache engine running on each EC2 instance that caches precompiled PHP scripts that boosts performance of PHP applications like Drupal. It is recommended to use a caching engine like OpCache when serving PHP pages for a website from Amazon EFS. OPcache can be configured to store it's cache in memory or on EBS volumes.
-
Mount the EFS file system using the default Linux mount options identified in the Amazon EFS User Guide. Please confirm that the following options are not used when mounting the EFS file system: actimeo=3 or acregmax=3 or acdirmax=3. These options generate significantly higher metadata operations by timing out the attribute caches more frequently.
-
Set the realpath_cache_size to atleast 512k. Also, please get the realpath_cache_size for your workload and make sure that it is less than 512k. You can do this by placing a php file (you can use any name – for example realpathcache.php) with the following contents in your Drupal directory and accessing. Please refresh the page multiple times before getting the final value:
<?php
print_r(realpath_cache_size());
?>
-
Please get the number of php files using “find . -type f -print | grep php | wc -l” in your Drupal directory. This number should be smaller than your opcache.max_accelerated_files settings. This setting controls how many PHP files, at most, can be held in memory at once. It's important that your project has LESS FILES than whatever you set this at.
-
Please set the opcache.memory consumption to 512MB (opcache.memory_consumption=512). The default value for this 64 (MB). In case memory size turns out to be a limiting factor for your workload, we can even configure opcache.file_cache. Also, please disable the opcache.validate_timestamps. Though it is not recommended that validate_timestamps is disabled in production, this ensures that calls are not being made to the NFS server to ensure opcache’s coherency during your testing.
To learn more about OPcache, please read http://php.net/manual/en/book.opcache.php
Drupal has a large partner ecosystem to further enhance the usability, performance, and ease of maintenance of Drupal deployments. Plugins allow you to leverage other AWS services like Amazon S3 and Amazon CloudFront to offload and store static content. Others may like the simplicity of storing all content on Amazon EFS and avoid installing and managing 3rd party plugins.
The master template receives all input parameters and passes them to the appropriate nested template which are executed in order based on conditions and dependencies. Review the template here aws-refarch-drupal-master.yaml
- Amazon Virtual Private Cloud (Amazon VPC)
- Internet Gateway (IGW)
- NAT Gateway (across all public subnets)
- Multiple Amazon VPC subnets (public & private) in 2 or 3 (if available) Availability Zones (AZs)
- Routing tables for public subnets - routing through IGW
- Routing tables for private subnets - routing through NAT Gateway
- Mulitple VPC Security Groups
- Bastion Auto Scaling Group (launching no instances) - in public subnets (public)
- Amazon Relational Database Service (Amazon RDS) Aurora cluster - in private subnets (data)
- Amazon Elastic File System (Amazon EFS) file system - with mount targets in private subnets (data)
- Amazon ElastiCache cache cluster (optional) - in private subnets (data)
- Amazon Elastic Load Balancing (Amazon ELB) Application Load Balancer (ALB) - in public subnets (public)
- Web Auto Scaling Group (launching 2 instances) - in private subnets (web)
- Amazon CloudFront distribution (optional)
- Amazon Route53 DNS record set (optional)
- EC2 Key Name Pair
- SSH Access CIDR block (to access bastion host)
- Database Name
- Database Master Username
- Database Master Password
- Database Size
- Database Instance Class Type
- Create ElastiCache cluster (boolean)
- ElastiCache Node Type
- Create CloudFront distribution (boolean)
- Create Route 53 record set (boolean)
- Web Instance Type
- CloudFront ACM certificate ARN (must be requested from us-east-1)
- ALB ACM certificate ARN (must be requested from the stack region using the Drupal site domain name)
- Drupal Title
- Drupal Administrator Username
- Drupal Administrator Username Password
- Drupal Administrator Email Address
- Drupal Site Domain Name (e.g. 'example.com')
- Drupal Main Language of the site
The master template receives all input parameters and passes them to the appropriate nested template which are executed in order based on dependencies. Review the template here aws-refarch-drupal-master.yaml
Review the template here aws-refarch-drupal-01-newvpc.yaml
The 'newvpc' stack creates the following network design:
Item | CIDR Range | Usable IPs | Description |
---|---|---|---|
VPC | 10.0.0.0/16 | 65,536 | The whole range used for the VPC and all subnets |
Web Subnet | 10.0.0.0/22 | 1022 | Private subnet in first Availability Zone |
Web Subnet | 10.0.4.0/22 | 1022 | Private subnet in second Availability Zone |
Web Subnet | 10.0.8.0/22 | 1022 | Private subnet in third Availability Zone (if available) |
Data Subnet | 10.0.12.0/22 | 1022 | Private subnet in first Availability Zone |
Data Subnet | 10.0.16.0/22 | 1022 | Private subnet in second Availability Zone |
Data Subnet | 10.0.20.0/22 | 1022 | Private subnet in third Availability Zone (if available) |
Public Subnet | 10.0.250.0/23 | 510 | Public subnet in first Availability Zone |
Public Subnet | 10.0.252.0/23 | 510 | Public subnet in second Availability Zone |
Public Subnet | 10.0.254.0/23 | 510 | Public subnet in third Availability Zone (if available) |
You can adjust the CIDR ranges used in this section of the aws-refarch-drupal-01-newvpc.yaml template:
Mappings:
SubnetConfig:
Vpc:
CIDR: 10.0.0.0/16
WebSubnet0:
CIDR: 10.0.0.0/22
WebSubnet1:
CIDR: 10.0.4.0/22
WebSubnet2:
CIDR: 10.0.8.0/22
DataSubnet0:
CIDR: 10.0.12.0/22
DataSubnet1:
CIDR: 10.0.16.0/22
DataSubnet2:
CIDR: 10.0.20.0/22
PublicSubnet0:
CIDR: 10.0.250.0/23
PublicSubnet1:
CIDR: 10.0.252.0/23
PublicSubnet2:
CIDR: 10.0.254.0/23
Review the template here aws-refarch-drupal-02-securitygroups.yaml
Review the template here aws-refarch-drupal-03-bastion.yaml
Review the template here aws-refarch-drupal-03-efs.yaml
Review the template here aws-refarch-drupal-03-elasticache.yaml
Review the template here aws-refarch-drupal-03-publicelb.yaml
Review the template here aws-refarch-drupal-03-rds.yaml
Review the template here aws-refarch-drupal-04-cloudfront.yaml
Review the template here aws-refarch-drupal-04-web.yaml
Review the template here aws-refarch-drupal-05-route53.yaml
If you found yourself wishing this set of frequently asked questions had an answer for a particular problem, please submit a pull request. The chances are that others will also benefit from having the answer listed here.
Portions copyright.
-
Drupal is licensed under the GNU General Public License (GPLv2 or later).
-
composer is licensed under the MIT License.
Please see LICENSE.txt for applicable license terms and NOTICE.txt for applicable notices.