Cloud computing services are replacing traditional hardware technologies at an extremely fast pace. The majority of businesses worldwide are already moving to the cloud or plan to in the near future. Over a short period of time, this technology took over the market as businesses preferred remote access to data as well as the cloud's scalability, economy, and reach.
Deploying to the cloud
COVID-19 and the resulting trend toward remote work forced organizations to adopt cloud technologies even if they hadn’t planned to originally. Software deployment to the cloud has also increased. Cloud is no longer just virtual machines, organizations are driving the use of Containers as a Service (CaaS) due to their growing interest in leveraging containers to ease development and testing, speed up deployment, scale operations, and increase the efficiency of workloads running in the cloud.
Since deployment to the cloud has become a standard practice, at GitLab we want to make this repeatable and boring. In this blog post, we explain how we've made deploying to Amazon Web Services (AWS) easier and we invite users to replicate this example to deploy to other cloud providers in a similar way.
Since we want cloud deployment to be as flexible as possible (similar to a microservices architecture), we constructed atomic Docker images that function as building blocks. Users can use these images as part of their custom
gitlab-ci.yml file or use our predefined
.gitlab-ci.yml templates. We also added the ability to use Auto DevOps with the new AWS deployment targets.
How to use GitLab's official AWS Docker Images
AWS CLI Docker image
In GitLab 12.6, we provided an official GitLab AWS cloud-deploy Docker image that downloads and installs the AWS CLI. This allows users to run
aws commands directly from their pipelines. For more information, see Run AWS commands from GitLab CI/CD.
Cloud Formation stack creation Docker image
In GitLab 13.5, we provided a Docker image that runs a script that creates a stack with Cloud Formation. The
gl-cloudprovision create-stack uses aws cloudformation create-stack behind the scenes. A JSON file based on the CloudFormation template must be passed to that command. For an example of this type of JSON file, see
cf_create_stack.json. With this type of JSON file, the command creates the infrastructure on AWS, including an EC2 instance directly from the
.gitlab-ci.yml file. The script exists once we get confirmation that the stack setup is complete or has failed (through periodic polling).
Push to S3 and Deploy to EC2 Docker image
In GitLab 13.5 we also provided a Docker image with Push to S3 and Deploy to EC2 scripts. The
gl-ec2 push-to-s3 script pushes code to an S3 bucket. For an example of the JSON file to pass to the
aws deploy push command, see
s3_push.json. This code can be whatever artifact is built from a preceding build job. The
gl-ec2 deploy-to-ec2 script uses
aws deploy create-deployment behind the scenes to create a deployment to an EC2 instance directly from the
.gitlab-ci.yml file. For an example of the JSON template to pass, see
create_deployment.json. The script ends once we get confirmation that the deployment has succeeded or failed (via polling).
Using GitLab CI templates to deploy to AWS
How to deploy to Elastic Container Service (ECS) with GitLab
In GitLab 12.9, we created a full
.gitlab-ci.yml template called
Deploy-ECS.giltab-ci.yml that deploys to ECS and extends support for Fargate. Users can include the template in their configuration, specify a few variables, and their application will be deployed and ready to go in no time. This template can be customized for your specific needs. For example: Replacing the selected container registry, changing the path of the file location, etc.
How to deploy to Elastic Cloud Compute (EC2) with GitLab
In GitLab 13.5, we created a full
.gitlab-ci.yml template called
CF-Provision-and-Deploy-EC2.gitlab-ci.yml that provisions the infrastructure by leveraging AWS CloudFormation. It then pushes your previously-built artifact to an AWS S3 bucket and deploys the pushed content to AWS EC2.
Security Considerations for AWS deployments
Predefined AWS CI/CD variables
In order to deploy to AWS, you must use AWS security keys to connect to to your AWS instance. Users can define this security keys as CI/CD environment variables that can be used by the deployment pipeline.
In GitLab 12.9, we added support for predefined AWS variables. This support function helps users know which variables are required for deploying to AWS and also prevents typos and spelling mistakes.
|Env. variable name||Value|
||Your Access key ID|
||Your Secret access key|
||Your region code|
"Just-in-time" guidance for AWS deployments
GitLab 13.1 provides just-in-time guidance for users who wish to deploy to AWS. Setting up AWS deployments isn't always as easy as we'd like it to be, so we've added in-product links to our AWS templates and documentation when you start adding AWS CI/CD variables to make it easier for you to use our AWS features. This will help you get up and running faster.
AWS guide from CI/CD variables
Added security for the GitLab's official AWS Docker images
In GitLab 13.5, we changed the image identifier from the release version number to the Docker image digest. Docker supports immutable image identifiers and we adopted this best practice to update our cloud-deploy images. When a new image is tagged, we also programmatically retrieve the image digest upon its build and create a release note to effectively communicate this digest to users. This guarantees that every instance of the service runs exactly the same code. You can roll back to an earlier version of the image, even if that version wasn't tagged (or is no longer tagged). This can even prevent race conditions if a new image is pushed while a deploy is in progress.
Docker image digest or release tag
Auto DevOps support for AWS deployments
GitLab already supports Kubernetes users deploying to AWS EKS cluster. Click the link to read instructions about how to deploy an application to a GitLab-managed Amazon EKS cluster with Auto DevOps.
We also expanded Auto DevOps to support non-Kubernetes users. Users can specify their deployment target by adding the
AUTO_DEVOPS_PLATFORM_TARGET variable under the CI/CD variables settings. Specifying the deployment target platform builds a full CI/CD pipeline that deploys to AWS targets.
We currently support:
AUTO_DEVOPS_PLATFORM_TARGET: ECS(added in GitLab 13.0)
AUTO_DEVOPS_PLATFORM_TARGET: FARGATE(added in GitLab 13.2)
AUTO_DEVOPS_PLATFORM_TARGET: EC2(added in GitLab 13.6)
For more information about Auto DevOps for AWS targets, see requirements for Auto DevOps documentation.
Here's a quick recording for how to use Auto Deploy to ECS:
Speed run on how to use auto deploy to EC2 (animation):
Future plans to extend AWS deployment support via GitLab
Check out some of the open issues below to see our plans are for the future of deploying to AWS using GitLab.
- Show AWS deployment success code in logs: This will bring the success/failure codes from AWS into your GitLab pipeline logs, allowing you to see the deployment success code without needing to go into the AWS console to retrieve the logs.
- Show AWS deployment success code in pipeline view: This will bring the success/failure codes from AWS into your GitLab pipeline, allowing you to see if the deployment job was successful in one view.
- Auto Deploy to AWS S3: This will expand the supported deployment targets covered in this blog to include S3 buckets as well.
- AWS integration per-environment role management: This returns a set of temporary security credentials you can use to access AWS resources that you normally might not be able to access. This is accomplished by using the AWS IAM roles.
More material on deploying to EKS and Lambda
We invite you to contribute to our other cloud provider solutions:
At GitLab, everyone can contribute. If you want to deploy to a target that isn't mentioned in this post, please let us know by adding an issue and linking it to our Natively support hypercloud deployments epic.