Argo

ArgoCD: "Argo CD is a declarative, GitOps continuous delivery tool for Kubernetes."

As to why Argo CD, they provide the following explanation: "Application definitions, configurations, and environments should be declarative and version controlled. Application deployment and lifecycle management should be automated, auditable, and easy to understand."

Simply said, ArgoCD is CD, not CI. We still need CI systems. Secondly, ArgoCD is running on Kubernetes, it's part of its ecosystem, as opposed to some other CI/CD systems. Finally, ArgoCD was built specifically for Kubernetes, not other platforms and systems.

Easier to explain the need for ArgoCD by direct comparison to another system that can do CD. Let's use Jenkins for this.

With Jenkins, you need make sure to install k8s related tools and set access for commands like kubectl. With ArgoCD you simply need to install it in your namespace but no need to install additional tools as it's part of k8s.

With Jenkins, managing access is usually done per pipeline and even if set globally in Jenkins, you still need to configure each pipeline to use that access configuration. With ArgoCD access management to k8s and other resources is given as it runs already on the cluster, in one or multiple namespaces.

With Jenkins, tracking the status of what got deployed to k8s can be done only as an extra step, by running the pipeline. This is because Jenkins isn't part of the k8s cluster. With ArgoCD you get much better tracking and visibility of what gets deployed as it runs in the same cluster and the same namespace.

With ArgoCD it's really easy to roll back to a previous version because all the changes done, are done to git which is a versioned source control. So it's enough to get to a previous commit for ArgoCD to detect a change and sync to the cluster. Worth to mention, this point specifically is true for Jenkins as well :)

1. A developer submitted change to an application repository
2. Jenkins pipeline is triggered to run CI on the change
3. If the Jenkins Pipeline completed successfully, build an image out of the new code
4. Push to image to a registry
5. Update K8S manifest file(s) in a separate app config repository
6. ArgoCD tracks changes in the app config repository. Since there was a change in the repository, it will apply the changes from the repo

False. It supports Kubernetes YAML files as well as Helm Charts and Kustomize.

It's the repository that holds app configuration, the one updated most of the time by CI/CD processes or DevOps, SRE engineers. In regards to ArgoCD it's the repository ArgoCD tracks for changes and apply them when they are detected.

• Your whole configuration is one place, defined as code so it's completely transparent, adjustable for changes and easily reproducible
• Everyone go through the same interface hence you have more people experiencing and testing the code, even if not intentionally
• Engineers can use it for testing, development, ... there is no more running manual commands and hoping to reach the same status as in the cluster/cloud.
• Single source of truth: you know that your GitOps is the repo from which changes can be done to the cluster. So even if someone tries to manually override it, it won't work.

Once Sorina made the modifications, ArgoCD will detect the state diverged and will sync the changes from the GitOps repository, overwriting the manual changes done by Sorina.

The answer is yes, it's possible. You can configure ArgoCD to sync to desired state when changes done manually and instead do something like sending alerts.

Imagine you have a cluster in the cloud, in one of the regions. Something happens to that cluster and it's either crashes or simply no longer operational.

If you have all your cluster configuration in a GitOps repository, ArgoCD can be pointed to that repository while be configured to use a new cluster you've set up and apply that configuration so your cluster is again up and running with the same status as o

Ella is right, ArgoCD is an extension of the cluster, that is very different from simply being deployed in the cluster as other CI/CD systems like Jenkins. ArgoCD uses existing k8s resources like K8s controllers (for monitoring and state differences) and etcd for storing data.

"Application"

It's a custom resource definitions which responsible for the deployment and synchronization of application resources to a Kubernetes cluster.

Instead of creating Kubernetes resources, you can use Git to manage who is allowed to push code, to review it, merge it, etc - either human users or 3rd party systems and services. There is no need to use ClusterRole or User resources in Kubernetes hence the management of access is much more simplified.

This section of an Application in ArgoCD, defines which Git repository should be synced

This section defines with which Kubernetes cluster the app in the tracked Git repository should be synced with.

AddProject

False. ArgoCD sync period is 3 minutes as of today (and not hours).

1. Gathers list of all the apps to sync (those that are marked with "auto-sync")
2. Gets Git state for each repository
3. Performs comparison between the repository Git state and the Kubernetes cluster state
4. If states are different, the application marked as "out-of-sync" and further action might be taken (based on the configuration)
5. If states are equal, the application marked as "synced"

Deploying applications in non-existing namespaces doesn't create the namespace. For that you have to explicitly mark "Auto-create namespace".

To fix it, you can simply run kubectl create namespace NAMESPACE_NAME but it's better of course to have it stored in Git rather than running kubectl commands.

argocd app create some-app \
--project  \
--repo https://fake.repo.address \
--path ./app_path \
--dest-namespace default \
--dest-server my.kubernetes.cluster

argocd app list

argocd app get some-app

argocd cluster add CLUSTER_ADDRESS/NAME

argocd cluster list

Yes, it is possible by adding the following to the argocd-cm (ConfigMap):

data:
  timeout.reconciliation: 300s

The value can be any number of seconds you would like to set.

sync functionality will be disabled.

A solution from Argo community in regards to managing multiple similar applications.

Basically a pattern where you have root application that consists of other child applications.

So instead of creating multiple separate applications, you have the root application pointing to a repository with additional applications.

• Cluster Preparation: You would like to deploy multiple applications at once to bootstrap a Kubernetes cluster

• Multiple environments: If deploying many versions of the same application, but with minor changes. For example, several test deployments to test different features

• Multiple clusters: when the same application needs to be deployed across multiple K8s clusters connected to ArgoCD

False, it can be deployed on one of them. ArgoCD is able to manage external clusters on which it doesn't run.

You don't usually want to go and update all of your clusters at once, especially when some for testing and development purposes and some for actual production usage.

There are multiple ways to deal with it:

1. Branch driven: Have branches for your GitOps repo where you push first to development, do some testing, merge it then to staging and if everything works fine in staging, you merge it to production.

2. Use overlays and Kustomize to control the context of where your changes synced based on the CI process/pipeline used.

• Healthy
• Missing: resource doesn't exist in the cluster
• Suspended: resource is paused
• Progressing: resources isn't healthy but will become healthy or has the chance to become healthy
• Degraded: resource isn't healthy
• Unknown: it's not known what's the app health

Not exactly. A Deployment (as well as StatefulSet, ReplicaSet and DaemonSet) considered healthy if the desired state equals to actual/current state (this includes the number of replicas).

True.

The health of custom Kubernetes resources is defined by writing Lua scripts.

You find such list of scripts here: https://github.com/argoproj/argo-cd/tree/master/resource_customizations

Automatic syncs means that once ArgoCD detected a change or a new version of your app in Git, it will apply the changes so the current/actual state can be equal to desired state.

With manual syncs, ArgoCD will identify there is a difference, but will do nothing to correct it.

If enabled, auto-pruning will remove resources when files or content is removed from a tracked Git repository.

If disabled, ArgoCD will not remove anything, even when content or files are removed.

Self-heal is the process of correcting the cluster state based on the desired state, when someone makes manual changes to the cluster.

ArgoCD is able to track packaged Helm chart in a sense where it will monitor for new versions.

True. Trying to execute commands like helm ls will fail because helm metadata doesn't exist anymore and the application is tracked as ArgoCD app.

A controller for Kubernetes to perform application deployments using different strategies like Blue/Green deployments, Canary deployments, etc.

In addition, it supports A/B tests, automatic rollbacks and integrated metric analysis.

• Argo Rollouts creates a new replicaset (that is the new app version)
  • Old version is still alive
• ArgoCD marks the app as out-of-sync

False. Quite common misconception today but both cab be used independency even though they work nicely together.

Shift towards fully automated rollbacks. Argo Rollouts supports multiple metric providers (Datadog, NewRelic, etc.) so you can use data and metrics for automating the rollbacks based on different conditions

Analysis is a resource deployed along a Rollout resources and defines the conditions and metrics threshols for performing a rollback

It's an Analysis resource that fetches response status from Prometheus (monitoring instance). If it's more than 0.90 the rollout will continue, if it's less than 0.90 a rollback will be performed meaning the canary deployment failed.

kubectl argo rollouts list rollouts

kubectl argo rollouts get rollout SOME-APP

kubectl argo rollouts status SOME-APP

kubectl argo rollouts set image SOME-APP web-app=some/registry/and/image:v2.0

kubectl argo rollouts promote SOME-APP

kubectl argo rollouts get rollout SOME-APP --watch