GitOps - Feat. Flux, Argo CD, and kapp-controller (You Choose!, Ch. 2, Ep. 2)

Choose Your Own Adventure: The Treacherous Trek to Production

In this episode, we'll figure out how to reconcile the actual state (Kubernetes resources) into the desired state stored in Git (Kubernetes manifests). This time, the choice will be between Flux, Argo CD, and Carvel kapp-controller.

#argocd #fluxcd #carvel #kubernetes




▬▬▬▬▬▬ 🔗 Additional Info 🔗 ▬▬▬▬▬▬

[powersurge]

What I am really interested to see is the Gitops for combined resources, that includes CSP resource provisioning that is back referenced in the helm chart (akv, managed identity, storage accounts, etc.)
This can be a challenging task, depending on the conditional approvals you may have on changes that are only visible at deployment time (e.g. terraform plan based approvals).
Another one, resource replacement if certain property is changed in terraform. Surely there are more scenarios that you don't want to blindly deploy. You need some CI health check logic during and after deployment. Focus still should be - do not introduce outage!

[mirceanton]

Thanks for answering my question about Flux/ArgoCD for admins(SRE) vs developers!

I had a discussion about this with a coworker a few days back and those were some really good points.
We were focusing mainly on the point Scott made, as in we considered Flux as something that just works under the hood and Argo as something that has more user-interaction due to the built-in UI.

[BernhardRode]

You talked about kickstarting a "new" account. This would be a great episode, too =)

[powersurge]

Something I didn't hear mentioned are the rollbacks (Helm atomic upgrade equivalent) which Flux supports. It is handy for self-service and allows for OPS-free deployment recoveries.