Solutions / Continuous delivery

Releases that are boring, auditable and reversible.

Every change goes through the same path: reviewed, tested, rolled out gradually and undone in minutes if something looks wrong.

What you get.

  • Small releases, many times a day
  • Rollback in one step
  • A record of who changed what, and when

For: Teams that still fear Friday deploys or depend on one person to release.

Releases

Deploy frequency14 / day
Change failure rate2%
Time to roll back1 min

How we do it.

CI pipelines.

Fast, secure pipelines that are the same for every team.

  • GitHub Actions, GitLab CI, Jenkins or Azure DevOps
  • OIDC to the cloud instead of stored keys
  • Image and IaC scans in every pull request
GitHub ActionsGitLab CIJenkinsAzure DevOpsOIDCTrivy

Pipeline

build1m 12s
scan0 critical
deploy3m 40s
Secrets in CInone

GitOps with ArgoCD.

Git is the only way into the cluster. Every change is a reviewed commit, and undoing it is a revert.

  • Promotion between environments through pull requests
  • Drift detected and corrected automatically
  • Multi-cluster delivery from one place
ArgoCDHelmKustomizeGit

Environments

devv2.14.0
stagingv2.14.0
prodv2.13.2 → PR #612

Progressive delivery.

New versions reach a small share of users first, and roll back on their own if error rates rise.

  • Canary and blue/green releases
  • Traffic shifting by percentage with Istio
  • Automatic rollback on SLO breach
IstioArgoCDPrometheus

Canary

v210%
Error rate0.02%
Next step25% in 15 min

Other solutions.