Skip to content

Argo CD

We now have an easy way to deploy and update a complex project using helm and helmfile but we still have to perform these actions manually! How can we automate this and follow GitOps principles? Enter Argo CD, a declarative, GitOps continuous delivery tool for Kubernetes.

Argo CD overview

Argo CD follows the GitOps pattern of using Git repositories as the source of truth for defining the desired application state. Kubernetes manifests can be specified in several ways: kustomize applications, helm charts, plain YAML manifests, and more. It then automates the deployment of the desired application states in the specified target environments by watching the changes in your git repo (via webhooks) and applying them in a continuous, self-healing manner.

Application deployments can track updates to branches, tags, or pinned to a specific version of manifests at a Git commit.

Argo CD needs to be installed on a Kubernetes cluster, and can from there drive other clusters (multi-clusters support). This is a very powerful pattern that is used by many organizations to manage complex infrastructures.


Another great contender of Argo CD is Flux CD.

Argo CD and helmfile

There is sadly no built-in support for helmfile, despite an open issue from August 2019.

One way to keep using helmfile is to have a CI workflow that uses helmfile template and saves the output as raw YAML manifests that can be later used by Argo CD. Another way is to use plugins. I found and tested two of them:

  • travisghansen/argo-cd-helmfile: this plugin is referenced directly in the open issue. It works well and has tons of options, but there is currently a bug in the Argo CD interface that prevents you from creating an AppProject directly from the UI using helmfile. All works great if you use the argocd CLI though.
  • lucj/argocd-helmfile-plugin: this plugin is fairly basic but works.

Installing Argo CD

Argo CD can be installed using helm or plain Manifests. Since we are using plugins and are now familiar with helmfile, let's install it this way:

  - name: argo

  - name: argocd
    namespace: argocd
      app: argocd
    chart: argo/argo-cd
    version: ~5.28.2
      - repoServer:
            - name: plugin
              image: lucj/argocd-plugin-helmfile:v0.0.11
              command: ["/var/run/argocd/argocd-cmp-server"]
                runAsNonRoot: true
                runAsUser: 999
                - mountPath: /var/run/argocd
                  name: var-files
                - mountPath: /home/argocd/cmp-server/plugins
                  name: plugins

Install it using:

helmfile -f argo-cd/install/helmfile.yaml sync 

Now that ArgoCD is installed, connect to the UI by creating a port-forward:

kubectl port-forward -n argocd svc/argocd-server 7000:80

ArgoCD is now available at https://localhost:7000.

The initial admin password is stored in a secret called argocd-initial-admin-secret in the argocd namespace. You can get it easily by executing:

kubectl -n argocd get secret argocd-initial-admin-secret -o jsonpath="{.data.password}" | base64 -d

Once you log in using the admin user and the password above, ensure you:

  1. delete the secret (kubectl -n argocd delete secret argocd-initial-admin-secret),
  2. change the admin password by clicking on the User Info in the left menu bar, then Update Password.

Basics of Argo CD

Argo CD registers new Kubernetes Custom Resources (CRs) that define new resource kinds. Those can be managed using kubectl, as regular built-in resources.

Argo CD has three constructs, defined in three different custom resource definitions or Kubernetes objects:

  • An AppProject is a high-level grouping mechanism in Argo CD that allows defining access controls and configuration overrides for a set of related applications.
  • An Application is a Kubernetes manifest and a set of parameters that define how a specific application should be deployed and managed by Argo CD. It defines a source repository, the tool used (helm, etc), the destination cluster, etc.
  • An ApplicationSet is a higher-level concept that allows defining multiple Applications with similar characteristics using templates, parameters and sources. It enables scaling the number of Application instances with variations in parameters, labels, and annotations. It can, for example, be used to spawn temporary environments when a pull request is created.

Note that you can create an app that creates other apps, which in turn can create other apps. This allows you to declaratively manage a group of apps that can be deployed and configured in concert.

Viewing Custom Resources

If you are curious, you can see the content of a CRD (Custom Resource Definition) using:

kubectl get crd | grep argo                                 2023-04-18T12:27:11Z                                2023-04-18T12:27:11Z                             2023-04-18T12:27:11Z
kubectl get crd -o yaml

You also query the instances of each crd using kubectl get. To discover the name you must use:

kubectl api-resources | grep argo 
applications     app,apps  true  Application
applicationsets  appset,appsets  true  ApplicationSet
appprojects      appproj,appprojs  true  AppProject
Hence, for applications, use kubectl get apps.

To create one of those resources, we can:

  • Use the Argo CD UI,
  • Use the argocd CLI, or
  • Use a YAML file and kubectl apply.

The App Project

By default, argocd creates a default App Project that has no restriction. Using default, you can deploy anything from and to anywhere.

How to see the default project definition

Use your Kubernetes knowledge!

kubectl get appprojects -n argocd default -o yaml
kind: AppProject
  creationTimestamp: "2023-04-18T12:27:34Z"
  generation: 1
  name: default
  namespace: argocd
  resourceVersion: "362197"
  uid: 2c95d565-7b55-4f60-a253-002704b82b15
  - group: '*'
    kind: '*'
  - namespace: '*'
    server: '*'
  - '*'
status: {}

Usually, we want to tighten security. For this demo, we will use the following AppProject. It creates a project called demo that limits the sources to one of my GitHub repositories ( and only allows certain kinds of resources to be created. The two namespaces kube-system and argocd are also restricted: they cannot be used by an app as a destination.

kind: AppProject
  name: demo
  namespace: argocd
  # Finalizer that ensures that project is not deleted until it is not referenced by any application
  description: Demo project to host rickroller
  # Allow manifests to deploy from any Git repos that are mine
    - "*"
    # Do not allow any app to be installed in `kube-system` or `argocd`
    - namespace: '!kube-system'
      server: "*"
    - namespace: '!argocd'
      server: '*'
    # Other destinations are fine though
    - namespace: '*'
      server: '*'
  # Limit what resources an app is allowed to create
    - group: ""
      kind: Namespace
    - group: ""
      kind: Pod
    - group: apps
      kind: ReplicaSet
    - group: apps
      kind: Deployment
    - group: apps
      kind: StatefulSet
    - group: ""
      kind: Service
    - group:
      kind: Ingress
    - group: ""
      kind: ConfigMap
    - group: ""
      kind: Secret
    - group: ""
      kind: ServiceAccount
    - group: ""
      kind: Role
    - group: ""
      kind: RoleBinding
    - group: ""
      kind: PersistentVolume
    - group: ""
      kind: PersistentVolumeClaim

Create the project using kubectl apply. You should be able to now see the project in the Argo CD UI, under Settings > Projects.

The App

Now that we have a project, we can create an app to deploy rickroller. Defining an app is quite straightforward:

kind: Application
  namespace: argocd
  name: rickroller
  # Perform a cascading delete of the Application's resources when the App is deleted
  project: demo # you could also use the "default" project if you want
    path: helmfile       # directory of the helmfile.yaml
    targetRevision: HEAD # use the default branch of the repo
    namespace: rickroller
    server: https://kubernetes.default.svc
    # Make Argo CD automatically apply changes without human intervention
      prune: true    # delete resources when they are no longer in git
      selfHeal: true # perform a sync when the cluster deviates from the state defined in Git (e.g. manual changes)
      - CreateNamespace=true # required to also create namespace

Create the app using kubectl apply. You should now see the app under Applications. Watch how Argo CD syncs all resources and deploys rickroller automatically! Isn't it great?

Using the argo CLI

You can perform the same operations (and more, such as syncing) using the argocd CLI. As an example, here is how you could create the rickroller App from the terminal:

#!/usr/bin/env bash

# This is equivalent to running kubectl apply -f app-rickroller.yaml
# It is just an example on how to use the argocd CLI
argocd app create rickroller \
    --project demo \
    --repo --path helmfile \
    --dest-server https://kubernetes.default.svc --dest-namespace rickroller \
    --set-finalizer \
    --sync-option CreateNamespace=true \
    --sync-policy automated --self-heal --auto-prune \

What now?

Now that we have our Argo CD set, we can commit and push any change and wait for the magic to happen. Since the app is configured with sync enabled, you should see your changes applied after around three minutes of committing. Why three minutes? Because Argo CD polls git repositories every three minutes to detect changes to the manifests.

The automatic sync interval is determined by the timeout.reconciliation value in the argocd-cm ConfigMap, which defaults to 180s (3 minutes).

If you want Argo CD to be more reactive, you will need to set up a webhook. This, of course, assumes your Argo CD instance is reachable from the outside.