Terraform deploy system

Terraform is an infrastructure-as-code tool that allows you to define both cloud and on-prem resources in human-readable configuration files that you can version, reuse, and share. CiviForm provides Terraform configuration files that allow you to deploy CiviForm on AWS. Knowledge of Terraform is not required to run them, but reading the high-level Terraform overview may be useful.

More detailed information on our deploy system can be found in our developer docs.

Setup

Outside configuration

You will need some values that are configured outside of CiviForm before you start the setup. Some of the steps are optional, meaning that you can bring up a staging environment and get the app working without them, but they will need to be completed for a production setup.

• (Optional) Admin auth client_id, client_secret, and discovery_uri. See Authentication Providers for more details on your type of provider.

• (Optional) Applicant auth client_id, client_secret, and discovery_uri. See Authentication Providers for more details on your type of provider.

• Domain name for your deployment. For example civiform.mycity.gov

• ARN of an SSL certificate for load balancer. See requesting AWS certificate

• Decision on how to organize your different CiviForm prod/staging/test instances. See AWS deployment setup options

Steps to run

1. Fork the civiform-deploy repo to your organization via the GitHub webpage.

2. Clone the repo onto the machine you are deploying from.

3. Find the version that you want to deploy on Github.

4. Copy the civiform_config.example.sh into civiform_config.sh and fill out the missing values. You should set all of the values appropriately for your environment, but the most important values are:

• CIVIFORM_MODE: Set to staging for all instances except for your production instance which should be prod.

• CIVIFORM_VERSION: Set this to the version string of the release you would like to deploy from Github, e.g. v3.28.0. Do not use latest, as this will be the latest development version and may be unstable.

• CIVIFORM_CLOUD_DEPLOYMENT_VERSION: This is the version of the cloud-deploy-infra repo to check out to perform the deployment. This repo is tagged with the same version as CiviForm releases so that you can be sure you are using the version of the deployment system that was fully validated working with that version. This should always be set to CIVIFORM_CLOUD_DEPLOYMENT_VERSION="${CIVIFORM_VERSION}" unless directed otherwise by CiviForm staff to deploy a particular fix outside of a release.

• CIVIFORM_APPLICANT_AUTH_PROTOCOL: Set to oidc or saml depending on your auth provider.

• CIVIC_ENTITY_LOGO_WITH_NAME_URL, CIVIC_ENTITY_SMALL_LOGO_URL, and FAVICON_URL should point to publicly accessible files used to brand your instance.

• BASE_URL: The URL of your civiform instance.

• AWS_REGION: Set this to the region closest to your jurisdiction for best performance.

• APP_PREFIX: If deploying multiple instances into the same AWS account (i.e. not separate accounts within an AWS Organization), they will each need to have a different app prefix so their resources do not conflict.

• SSL_CERTIFICATE_ARN: This should be copied from the information for the SSL certificate created in AWS that is verified with your DNS provider.

• The values in the generic-oidc and/or ADFS and Azure AD configuration sections according to your auth provider.

1. Run bin/doctor and install the dependencies.

2. Run bin/setup. What to expect:

   • Takes up to 20 minutes to run. Make sure you have time to allow the script to run to completion to avoid errors.

   • Terraform brings up resources in the cloud (database, network, server, etc).

   • Asks confirmation a few times before creating resources, listing everything that will be created.

   • Safe to re-run the script if it fails (re-runs will take longer because resources must be destroyed before being re-created).

The configuration values in civiform_config.sh represent the desired state of your CiviForm deployment. The bin/setup and bin/deploy commands work to make your cloud environment match the desired state. If a command fails, your cloud environment may not match the desired state. These commands are safe to retry if they fail. If a command is persistently failing, you can work with our on-call engineer to resolve the issue. Our on-call engineer responds to new issues in the CiviForm issue tracker (see the on-call guide for details).

Note: If you are running bin/doctor or another command for a config file other than civiform_config.sh, you can specify that with the config flag (i.e. bin/doctor --config=civiform_staging_config.sh)

Deploy

1. Find the version that you want to deploy on Github.

2. Update the CIVIFORM_VERSION value in civiform_config.sh.

3. Run bin/deploy.

Rotating the application secret

The Play Framework used by CiviForm utilizes an application secret key. This key is used for signing session cookies and CSRF tokens, among other things. While this secret is stored securely by the deployment system, it's a good idea to rotate it periodically in order to mitigate the risk of a leaked secret.

IMPORTANT: When the secret is rotated, all user sessions will be invalidated. This means that any guest users in the middle of an application that have not submitted it yet will lose that application, and any logged in users or admins will get logged out. It is recommended that this rotation happen at a low traffic time of day. You may also want to give users a warning before it happens.

To rotate the secret, run bin/run, and enter the rotate_app_secret command. This will redeploy CiviForm, changing the secret key to a new, random value.

Additionally, you should add export RANDOM_PASSWORD_LENGTH=64 to your civiform_config.sh file. The secret length was originally only 16 characters. CiviForm uses Play Framework 3.x, which requires a minimum length of 32 characters.

Troubleshooting

Terraform fails with error "Provider produced inconsistent final plan"

This error can happen when running bin/setup for the first time. If you see it, re-run bin/setup. This is a known Terraform bug.

Terraform fails with other errors

The deploy command is idempotent, so if it fails, try running it again. The setup command can also be re-run, but it may fail to destroy resources if the resources Terraform uses to track created resources are corrupted. If the script is failing to destroy resources, try running bin/run destroy on its own. If this fails, you may need to manually delete resources in your cloud provider's console. Once all resources are cleaned up, run bin/run destroy_backend_state_resources to cleanup up the corrupted backend state resources and then try running bin/setup again.

If changes were made upstream, you can change the code in the checkout env, but will need to commit PRs to fix in the main repo.

No such file or directory

If you see error like "no such file or directory"

./db-connection: line 2: cloud/aws/bin/lib.sh: No such file or directory
./db-connection: line 21: out::error: command not found

The scripts expect you to be in specific directories. You probably need to cd into the checkout directory or the top level directory. If you are running setup/deploy, you will need to be in the top level directory. If you are running a script like db-connection, you need to be in the checkout directory.

Terraform fails with Error acquiring the state lock

This situation can happen when exiting deployment scripts using "Ctrl-C". Terraform acquires a lock every time you run ./bin/deploy or ./bin/setup, and releases the lock at the end of the script. This helps to prevent concurrent infrastructure changes. If the deploy process exited outside of Terraform, the lock remains and needs to be force removed in order to run deploy again. The end of the error message will contain the LOCK_ID. To unlock, re-run either bin/deploy or bin/setup with FORCE_UNLOCK_ID set like this:

FORCE_UNLOCK_ID="$LOCK_ID" ./bin/deploy

Alternatively, it can be done by manually calling terraform like this:

terraform -chdir=checkout/cloud/aws/templates/aws_oidc force-unlock $LOCK_ID

(AWS) Notes on using CloudShell for deployment

It is an option to use CloudShell for doing deployments, but you should be aware that installations aren't persisted (see FAQs), so it may be easier to use the AWS CLI.

If you do choose to use CloudShell, you likely will need to install Java and Terraform.

Java can be installed with sudo yum install java and terraform can be installed by following the installation commands under Linux / Amazon Linux.

Helpful resources

Requesting AWS certificate

Follow official documentation. After your certificate is ready and validated, copy the ARN from your console and put it into the SSL_CERTIFICATE_ARN variable in your civiform_config.sh

ARN has format of arn:aws:acm::<user_id>:certificate/<identifier>

Screen Shot 2022-08-10 at 3 50 39 PM

AWS deployment setup options

Before going through the AWS deployment, it is helpful to understand how you plan to manage various deployments, if you plan to create multiple instances of CiviForm (i.e. staging, prod, etc.).

By default, each deployment will have its own prefix, which allows you to distinguish the different ones from each other, but AWS Organizations is a way of keeping deployments more separate from each other.

Some benefits of using AWS environments are:

• There is more separation between environments, which makes it easier to remove resources without worrying about causing any issues to another deployment environment

• It is easier to see cost breakdowns by deployment

A drawback to be aware of is that you'll have to create profiles in the AWS CLI and switch between the different profiles when doing deployments or, if using CloudShell, you'll have to do this in separate CloudShell for each environment.

Access the database for emergency repair

The CiviForm deployment system provides a mechanism for temporary and secure direct access to the production database via the pgadmin

For details on how to access the database with pgadmin visit the Production Database Access page.