Removing kubectl from CI Pipelines: A Slower, More Honest Deployment Model

Valerio Uberti — Mon, 27 Jul 2026 13:11:30 +0000

Most teams adopt ArgoCD to speed up Kubernetes deployments. The real value is that it slows them down. The shift from push-based CI/CD to pull-based GitOps introduces a 45–60 second reconciliation lag — and eliminates an entire category of incidents.

The full breakdown (with code, benchmarks, and the culture gap every platform team hits) is on my blog:

👉 Read the complete article here

Here is the thesis: a CI pipeline should not touch the cluster. It should touch Git. ArgoCD, running inside the cluster, should be the only component that applies manifests.

The counter-intuitive part: this adds latency. A 50-second sync lag feels like a regression. But what teams lose in speed, they gain in honesty automatic drift detection, rollbacks as git revert, and zero cluster credentials in CI.

Consider an internal platform serving ~80 engineers. Before GitOps, the platform team fields ~8 support tickets per month just answering "what is actually running?" After the move, that number drops to zero. The trade-off is correct.

If a platform team is still running kubectl apply from GitHub Actions, this is worth reading.

Tags: #kubernetes #gitops #argocd #devops #platform-engineering

DEV Community: Valerio Uberti

Removing kubectl from CI Pipelines: A Slower, More Honest Deployment Model