Skip to content

GitOps

Gitops is a way of implementing Continuous Deployment for containerized applications.

The core idea of GitOps is having a Git repository that always contains declarative descriptions of the infrastructure currently desired in the production environment and an automated process to make the production environment matches the described state in the repository.

From the infrastructure point of view, we want to control the governance of the infrastructure and be sure that what is expected, is what happened.

Requirements

Developer and operation teams want to:

  • Audit all changes made to pipelines, infrastructure, and application configuration.
  • Roll forward/backward to desired state in case of issues.
  • Consistently configure all deployment environments.
  • Reduce manual effort by automating application and environment setup and remediation.
  • Have an easy way to manage application and infrastructure state across clusters/environments

GitOps is a natural evolution of DevOps and Infrastructure-as-Code.

Core principles

  • Git is the source of truth for both application code, application configuration, dependant service/product deployments, infrastructure config and deployment.
  • Separate application source code (Java/Go) from deployment manifests i.e the application source code and the GitOps configuration reside in separate git repositories.
  • Deployment manifests are standard Kubernetes (k8s) manifests i.e Kubernetes manifests in the GitOps repository can be simply applied with nothing more than a oc apply or kubectl apply.
  • Kustomize.io is used as template engine for defining the differences between environments i.e reusable parameters with extra resources described using kustomization.yaml.
  • Minimize yaml duplication - no copy/paste
  • Support two axis of configuration: clusters and environments: prod, test, dev. (Accept separating production repo if organization is willing to do so).
  • Prefer a multi-folder and/or multi-repo structure over multi-branch. Avoid dev or test configuration in different branches. In a microservices world, a one branch per environment will quickly lead to an explosion of branches which again becomes difficult and cumbersome to maintain.
  • Minimize specific gitops tool dependencies: (Try to converge to Tekton and ArgoCD)

  • Do not put independent applications or applications managed by different teams in the same repo.

  • Day 1 Operations are actions that users take to bootstrap a GitOps configuration and on how to set up GitOps and Sealed Secrets.

  • Day 2 Operations are actions that users take to change a GitOps system.

Using continuous delivery approach and tool like ArgoCD, the deployment of all the components is controlled by the tool from the Gitops repositories.

Supporting tools

Git Action

Git Action

Configure github SECRETS for each project:

  • DOCKER_REGISTRY quay.io
  • DOCKER_REPOSITORY jbcodeforce
  • DOCKER_USERNAME jbcodeforce
  • DOCKER_PASSWORD
  • DOCKER_IMAGE_NAME the name of the application

Define a .github/workflow/dockerbuild.yaml to declare how to build the app.

OpenShift GitOps

OpenShift GitOps is the operator and configuration to run ArgoCD, as a controller inside OpenShift.

The environment configuration repository defines the desired state of the application/solution.

Using OpenShift console > Operator Hub, search for GitOps and select Red Hat OpenShift GitOps operator:

Then click install and use the default parameters.

To do it via oc CLI, use the eda-gitops-catalog gitops repository and the command:

oc apply -k openshift-gitops/operator/overlays/

It can take some minutes to get it installed.

oc describe operator openshift-gitops-operator

Proposed git repository structure

There are different ways to organize projects: The KAM tool proposes to create one gitops repository to control the configuration and deployment of each services and apps of the solution.

Another solution is to use a three repositories structure, that will match team structure and persona: dev, SRE.

  • application: deployment.yaml, config map... for each application. Developers lead this repository.
  • shared, reusable services like Kafka, Database, LDAP,... as reusable services between environments: Dev and operations ownership
  • Cluster and infrastructure: network, cluster, storage, policies... owned by operation

With this three level structure, each "solution" will have 3 separate repositories:

  • solution_name-gitops-apps
  • solution_name-gitops-services
  • solution_name-gitops-infra

Now the different deployment environments can be using different k8s clusters or the same cluster with different namespaces.

With the adoption of ArgoCD, developer can have one bootstrap app that starts other apps and monitor each of those layers.

See detail in this separate note.

OpenShift Projects

For each solution we may use a number of different projects:

  • one solutionname-cicd for Tekton pipeline run...
  • one solutionname-dev for the application instances in a development mode.
  • one solutionname-staging for pre-production tests
  • one solutionname-prod for production deployment. This can be in the same OCP cluster or in different one.

Within the -dev project we can isolate the dependant applications, like Kafka cluster, or postgresql, where with Staging and production we can use multi-tenant deployment of such applications.

The KAM tool supports this approach.

Services

For the Service level, try to adopt catalog repositories to hold common elements that will be re-used across teams and across other repositories. See example in Red Hat Canada Catalog git repo and our Business automation or EDA catalogs.

Use Kustomization to reference remote repositories as a base and then patch it as needed to meet your specific requirements.

bases:
 - github.com/ibm-cloud-architecture/eda-gitops-catalog/kafka-strimzi/operator/overlay/stable?ref=main
 - ../base

Reference the common repository via a tag or commit ID.

Application

For Application repositories, align your repositories along application and team boundaries: one gitops repo per application, different folders for microservices and deployable components.

Use bootstrap folder to load the application components into the cluster: this will be an ArdoCD app.

Use Overlay for the different app configuration depending of the target environment.

Dedicated a folder (tekton) for pipelines, tasks.

If you find you are duplicating argoCD manifests in the bootstrap folder, then create a separate high level argocd folder.

Process

A key question in any gitops scenario is, how to manage promotion of changes in manifests between different environments and clusters.

Some principles to observe:

  • Every change in the manifests needs to flow through the environments in hierarchical order, i.e. (dev > test > prod)
  • We do not want a change to the base flowing to all environments simultaneously
  • Tie specific environments to specific revisions so that changes in the repo can be promoted in a controlled manner We can manage the revision in the gitops tools or in kustomize or both.

The pre and post synch processing

When Argo CD starts a sync, it orders the resources in the following precedence:

  • The phase (See resource hook doc)

  • The wave they are in (lower values first). Defined as annotations in the different resources 

    metadata:
  annotations:
    argocd.argoproj.io/sync-wave: "5"

  • By kind (e.g. namespaces first)

  • By name

It then determines the number of the next wave to apply. This is the first number where any resource is out-of-sync or unhealthy.

OpenShift Pipelines

OpenShift Pipelines is a Continuous Integration / Continuous Delivery (CI/CD) solution based on the open source Tekton project. The key objective of Tekton is to enable development teams to quickly create pipelines of activity from simple, repeatable steps. A unique characteristic of Tekton that differentiates it from previous CI/CD solutions is that Tekton steps execute within a container that is specifically created just for that task.

Users can interact with OpenShift Pipelines using the web user interface, command line interface, and via a Visual Studio Code editor plugin. The command line access is a mixture of the OpenShift oc command line utility and the tkn command line for specific Tekton commands. The tkn and oc command line utilities can be downloaded from the OpenShift console web user interface. To do this, simply press the white circle containing a black question mark near your name on the top right corner and then select Command Line Tools:

CLI tools

See more detail in this Tekton summary

Kustomize and gitops

There are two ways to implement the deployment strategy for GitOps:

  • Push-based: use CI/CD tools like Jenkins, Tekton, Travis... to define a pipeline, triggered when application code source is updated. Pipelines build the container image and deploy the modified yaml files to the GitOps repo. Changes to the environment configuration repository trigger the deployment pipeline. It has to be used for running an automated provisioning of cloud infrastructure. See also this k8s engine 'cloud build' tutorial.
  • Pull-based: An operator takes over the role of the pipeline by continuously comparing the desired state in the environment repository with the actual state in the deployed infrastructure.

A CI/CD based on git action will build the image and edit Kustomize patch to bump the expected container tag using the new docker image tag, then will commit this changes to the gitops repo.

Kustomize is used to simplify the configuration of application and environment. Kustomize traverses a Kubernetes manifest to add, remove or update configuration options without forking. It is available both as a standalone binary and as a native feature of Kubectl. A lot of Kustomize examples here..

The simple way to organize the configuration is to use one kustomize folder, then one folder per component, then one overlay folder in which environment folder include kustomization.yaml file with patches.

└── postgres
    ├── base
    │   ├── configmap.yaml
    │   ├── kustomization.yaml
    │   ├── pvc.yaml
    │   ├── secret.yaml
    │   ├── service-account.yaml
    │   ├── statefulset.yaml
    │   ├── svc-headless.yaml
    │   └── svc.yaml
    ├── kustomization.yaml
    └── overlays
        └── dev
            ├── kustomization.yaml
            └── sealedsecret.yaml

Here is an example of kustomization.yaml:

bases:
  - ../../base
patchesStrategicMerge:
  - ./secret.yaml

The patchesStrategicMerge lists the resource configuration YAML files that you want to merge to the base kustomization. You must also add these files to the same repo as the kustomization file, such as overlay/dev. These resource configuration files can contain small changes that are merged to the base configuration files of the same name as a patch.

In GitOps, the pipeline does not finish with something like oc apply... but it’s an external tool (Argo CD or Flux) that detects the drift in the Git repository and will run these commands.

Future readings

KAM - Gitops Application Manager CLI

KAM's goal is to help creating a GitOps project for an existing application as day 1 operations and then add more services as part of day 2 operation.

GitOps approach is to manage deployments into one or more environments (a namespace in a Kubernetes). One or more applications can be deployed into a given environment. An application is an aggregation of one or more services. The source code for each service (or microservice) is contained within a single Git repository

Pre-requisites

Have the OpenShift GitOps and OpenShift Pipelines operators installed.

Be connected to the cluster.

Creating a GitOps project for a given solution

Use the following kam command to create a gitops repository with multiple ArgoCD applications. To run successfully, we need to be connected to OpenShift cluster with Sealed Secret, GitOps and Pipelines operators deployed, get SSH key from github if repositories are used:

# list the  available options
kam bootstrap -- 
# create a gitops repo by specifying argument
kam bootstrap \
--service-repo-url https://github.com/jbcodeforce/refarch-eda-store-inventory \
--gitops-repo-url  https://github.com/jbcodeforce/rt-inventory-gitops \
--image-repo image-registry.openshift-image-registry.svc:5000/ibmcase/store-aggregator \
--output rt-inventory-gitops \
--git-host-access-token <a-github-token> \
--prefix rt-inventory --push-to-git=true

To retrieve the github access token see this note.

Kam bug: the application name is matching the git repo name, and there will be an issue while creating binding with the limit of the number of characters. kam bootstrap needs a --service-name argument.

This will create a GitOps project, push to github.com as private repo. The repo includes two main folders and a pipelines.yaml describing your first application, and the configuration for a complete CI pipeline and deployments from Argo CD.

At the same level of the folder hierarchy, there is a secrets folder to keep secrets to do not commit to Git. See secrets section below.

What is inside

Config

The config folder defines argocd and cicd Argo applications to support deployment and pipeline definitions:

  ├── config
  │   ├── argocd
  │   │   ├── argo-app.yaml
  │   │   ├── cicd-app.yaml
  │   │   ├── kustomization.yaml
  │   │   ├── rt-inventory-dev-app-refarch-eda-store-inventory-app.yaml
  │   │   ├── rt-inventory-dev-env-app.yaml
  │   │   └── rt-inventory-stage-env-app.yaml
  │   └── rt-inventory-cicd
  │       ├── base
  │       │   ├── 01-namespaces
  │       │   │   └── cicd-environment.yaml
  │       │   ├── 02-rolebindings
  │       │   │   ├── argocd-admin.yaml
  │       │   │   ├── pipeline-service-account.yaml
  │       │   │   ├── pipeline-service-role.yaml
  │       │   │   └── pipeline-service-rolebinding.yaml
  │       │   ├── 03-tasks
  │       │   │   ├── deploy-from-source-task.yaml
  │       │   │   └── set-commit-status-task.yaml
  │       │   ├── 04-pipelines
  │       │   │   ├── app-ci-pipeline.yaml
  │       │   │   └── ci-dryrun-from-push-pipeline.yaml
  │       │   ├── 05-bindings
  │       │   │   ├── github-push-binding.yaml
  │       │   │   ├── rt-inventory-dev-app-refarch-eda-store-simulator
                  └── refarch-eda-store-simulator-binding.yaml
  │       │   ├── 06-templates
  │       │   │   ├── app-ci-build-from-push-template.yaml
  │       │   │   └── ci-dryrun-from-push-template.yaml
  │       │   ├── 07-eventlisteners
  │       │   │   └── cicd-event-listener.yaml
  │       │   ├── 08-routes
  │       │   │   └── gitops-webhook-event-listener.yaml
  │       │   └── kustomization.yaml
  │       └── overlays
  │           └── kustomization.yaml

The major elements of this configuration are:

  • argo-app: root application to manage the other applications: It points to the kustomization.yaml under config/argocd so when adding new microservice to the solution using kam cli then this kustomize will be updated triggering a new argo app to be created
    • argo-app.yaml
    • cicd-app.yaml
    • rt-inventory-dev-app-refarch-eda-store-simulator-app.yaml
    • rt-inventory-dev-env-app.yaml
    • rt-inventory-stage-env-app.yaml
  • cicd-app: Argo app to monitor config/rt-inventory-cicd/overlays which defines -cicd namespace, pipeline service account, then role bindings. And it also defines Tekton tasks, pipelines, triggers... This is where developers can add new tasks and pipelines for their applications.
  • rt-inventory-dev-env-app: Argo app to monitor environments/rt-inventory-dev/env/overlays which defines the NameSpace, Service Account, RoleBindings, for a dev environment.
  • rt-inventory-dev-app-refarch-eda-store-simulator: Argo app to monitor environments/rt-inventory-dev/apps/app-refarch-eda-store-simulator/overlays to deploy a specific app

Environment

environment includes two environment definitions for dev and stage. For each environment the apps folder includes the deployment of all the apps part of the solution and then a env folder to define the dev namespace, role bindings... 

    environments
  │   ├── rt-inventory-dev
  │   │   ├── apps
  │   │   │   └── app-refarch-eda-store-inventory
                ..... SEE BELOW
  │   │   └── env
  │   │       ├── base
  │   │       │   ├── argocd-admin.yaml
  │   │       │   ├── kustomization.yaml
  │   │       │   ├── my-inventory-dev-environment.yaml
  │   │       │   └── my-inventory-dev-rolebinding.yaml
  │   │       └── overlays
  │   │           └── kustomization.yaml
  │   └── rt-inventory-stage
  │       └── env
  │           ├── base
  │           │   ├── argocd-admin.yaml
  │           │   ├── kustomization.yaml
  │           │   ├── rt-inventory-stage-environment.yaml
  │           │   
  │           └── overlays
  │               └── kustomization.yaml

In an apps folder the services/${app-name-here}/base/config defines the manifests to configure the application. This is where we can put you app specific configuration.

│   ├── rt-inventory-dev
│   │   ├── apps
│   │   │   └── app-refarch-eda-store-simulator
│   │   │       ├── base
│   │   │       │   └── kustomization.yaml
│   │   │       ├── kustomization.yaml
│   │   │       ├── overlays
│   │   │       │   └── kustomization.yaml
│   │   │       └── services
│   │   │           └── refarch-eda-store-simulator
│   │   │               ├── base
│   │   │               │   ├── config
│   │   │               │   │   ├── 100-deployment.yaml
│   │   │               │   │   ├── 200-service.yaml
│   │   │               │   │   ├── 300-route.yaml
│   │   │               │   │   └── kustomization.yaml
│   │   │               │   └── kustomization.yaml
│   │   │               ├── kustomization.yaml
│   │   │               └── overlays
│   │   │                   └── kustomization.yaml

Attention! if a service is used between multiple applications, ( for exmaple Kafka), then we need to add under the environments/....-dev/ folder a services folder to declare such services. To see an example of this approach go to this repo.

Secrets

At the same level as the GitOps folder created by kam, there is a secrets folder to be used to manage secrets without getting them into git.

The approach is to use Bitmani Sealed Secret operators to deploy a server on OpenShift to decrypt in the final format, encoded sealed secrets to a kubernetes secret into target namespace.

Sealed Secrets are a "one-way" encrypted Secret that can be created by anyone, but can only be decrypted by the controller running in the target cluster.

It needs kubeseal CLI on the client side to do encryption that only the controller can decrypt.

  • Install the Sealed Secret Operator in sealed-secrets project
# Under one of the catalog repo: eda or dba
oc apply -k sealed-secrets-operator/overlays/default
  • Create a Sealed secret controller: for example sealed-secrets-controller. The previous command will create it.
oc get pods -n sealed-secrets
NAME                                       READY   STATUS    RESTARTS   AGE
sealed-secrets-controller-8dd96b4d-4zq2b   1/1     Running   0          4m39s
  • Then for each secrets in secrets folder do something like:
cat gitops-webhook-secret.yaml| kubeseal --controller-namespace sealed-secrets \
--controller-name sealed-secrets-controller --format yaml >gitops-webhook-sealedsecret.yaml
#
cat git-host-access-token.yaml | kubeseal --controller-namespace sealed-secrets \
--controller-name sealed-secrets-controller --format yaml > git-host-access-token-sealedsecret.yaml
# 
cat docker-config.yaml | kubeseal --controller-namespace sealed-secrets \
--controller-name sealed-secrets-controller --format yaml > docker-config-sealedsecret.yaml 
# 
cat git-host-basic-auth-token.yaml | kubeseal --controller-namespace sealed-secrets \
--controller-name sealed-secrets-controller --format yaml > git-host-basic-auth-token-sealed.yaml
# then continue with any microservice webhook secret

Each sealed secret can be in uploaded to github, and applied to different k8s namespace.

# mv all the sealed.yaml to config/...-cicd/base/09-secrets folder
# in secrets folder
mv *-sealed.yaml ../*gitops/config/

oc apply -f gitops-webhook-sealedsecret.yaml
# this should create a k8s secret

ERROR: giving up: no key could decrypt secret (.dockerconfigjson)

What to do from there

  1. Assess if you need the Pipeline or not (for example when defining a demo environment with public docker images we do not need Pipeline). If you do not use Pipeline, comment in the config/argocd/kustomization.yaml the call to cicd-app.yaml.
  2. Assess if you need stage environment. For demo one environment is needed.
  3. Change the environments/<>-dev/apps/app-<appname>/services/<appname>/base/config file with the kustomization files of your application.

  4. Add any cross microservice/application dependent services under a new structure per environment. Ex

    environments/<>-dev/services/kafka-strimzi/

  5. Create a bootstrap folder and add the declaration for the ArgoCD project to avoid using default project. See example in this repo.

  6. Bring up the Argocd app of app by doing: oc apply -k config/argocd/, which will create:

    • three Argo applications to monitor CI/CD, Dev and Staging environments: rt-inventory-cicd, rt-inventory-dev, rt-inventory-stage
    • One Argo Application per service to deploy: the following figure has only one of such service.
    • Three OpensShift projects: one for cicd, one per target 'environments' (dev, stage)

    depending of the OCP version, there may be issue regarding controller user being able to create resource. To enable full cluster admin access on OpenShift, run the following command: 

    oc adm policy add-cluster-role-to-user cluster-admin -z argocd-application-co

  7. Part of the configuration bootstraps a simple OpenShift Pipelines pipeline for building code when a pull-request is opened.

  8. Add new microservices using command like:

    kam service add --git-repo-url https://github.com/jbcodefoce/new-service.git --app-name newservicename  --env-name stage  --image-repo quay.io/jbcodefoce/new-service --service-name new-service

    Attention there is an issue with KAM as it will overwrite any update in previously generated file, so for example bringing the project name back to default. We need to tunes the pipelines.yaml file

    For each service, an unencrypted secret will be generated into the secrets folder. Make sure to apply these secrets to the cluster.

  9. Add a webhook to each microservice repository to connect to the Pipeline event listener.

    # Get the event listener route end point
oc get route --namespace rt-inventory-cicd

    Then in the application github repository do the following:

    • Add a secret with the same password as the created by kam: something like webhook-secret-rt-inventory-dev-refarch-eda-store-simulator. 

     oc get secret  webhook-secret-rt-inventory-dev-refarch-eda-store-simulator -o jsonpath='{.data.webhook-secret-key}' | base64
    * use, > Settings > Webhook > add new webhook, copy this address with http:// prefix. Try to do a new push to see if the event is sent to the listener. If so a new pipeline is started visible in the Pipelines menu on OpenShift console.

Future readings