DevOps Roadmap 2026: 11 Tools That Actually Matter

DevOps Learning Roadmap for 2026

The video presents an ordered DevOps learning roadmap, prioritizing tools the presenter says appear in job postings and remain useful as AI increases the amount of code teams ship.

The central argument is that AI can amplify a team’s existing strengths and weaknesses. Faster delivery can also mean more failures, making sound operational fundamentals especially important.

Roadmap, in order

1. Linux, Git, and one programming language — Learn Python or Go. The presenter cites an analysis of 850,000 US job postings in which Git appeared more often than Kubernetes and Docker combined. They also note that entry-level hiring at major tech companies is down compared with 2019, as AI has absorbed some routine junior-level work.

2. Containers: Docker — Docker packages an application and its dependencies so it runs consistently across environments. The presenter cautions that Docker Desktop’s free-use eligibility depends on company size and revenue. They also warn that Docker Hub’s unauthenticated pull limits can break CI pipelines and recommend setting up a mirror.

3. Container orchestration: Kubernetes — Kubernetes runs containers across machines, restarts failed workloads, and scales them. The presenter recommends learning it, but says cloud-provider skills may be a better next priority if time is limited.

4. Kubernetes traffic routing: Gateway API — The presenter says Ingress NGINX stopped receiving security fixes and recommends checking whether it is in use and migrating to Gateway API. They mention a Kubernetes migration tool and a separate video covering migration methods. Their advice is to learn Gateway API itself, since implementations can vary.

5. Infrastructure as code: Terraform or OpenTofu — The presenter treats these as one transferable skill with two binaries: OpenTofu is the open-source fork, while job postings may still use the Terraform name. For interviews, they emphasize understanding Terraform state, locking, importing existing infrastructure, and appropriately scoped IAM policies—not just syntax.

6. CI/CD: GitHub Actions, plus Jenkins migration skills — The presenter recommends GitHub Actions and warns against pinning workflows to movable version tags. Instead, pin actions to a commit SHA, which identifies exact code. This advice follows a cited incident in which a compromised action exposed secrets in build logs. Jenkins, they argue, remains valuable mainly for maintaining or migrating existing systems.

7. GitOps: Argo CD — Git serves as the source of truth for what runs in the cluster, and manual changes are reconciled back to it. The presenter favors Argo CD partly for its interface, while stressing that a cited CNCF survey covered Argo CD and did not provide a comparable Argo-versus-Flux statistic.

8. Observability: OpenTelemetry, Prometheus, Grafana, and Alertmanager — The presenter groups metrics, logs, traces, and alerts as parts of one observability practice. OpenTelemetry can help teams instrument once and change monitoring backends later. They also warn that the open-source Grafana OnCall project was archived, removing its paging features; the replacement is cloud-only.

9. Security: Trivy and Kyverno — Trivy can scan images, infrastructure code, and repositories for vulnerabilities, misconfigurations, and leaked secrets. Kyverno lets teams write Kubernetes policies in YAML. The presenter suggests it as a simpler starting point for Kubernetes-only policy needs, while describing OPA as more powerful but requiring Rego. They also highlight the EU Cyber Resilience Act’s reporting and software bill of materials obligations, citing a September 11 deadline and a 10-year retention requirement.

10. Workflow orchestration: Kestra, after the fundamentals — The presenter considers orchestration a lower priority than Terraform and Kubernetes. Kestra uses YAML workflows, supports multiple languages, and has integrations for tools such as Terraform, Ansible, Argo CD, and Kubernetes. Caveats include its 2.0 release-candidate status and enterprise-only SSO and access control.

11. AI in operations: learn MCP, but keep humans in control — The presenter describes the Model Context Protocol (MCP) as a standard for AI tools to communicate with other tools, noting integrations in Kestra, Ansible, and Grafana. They advise using AI to help read logs, not letting it autonomously execute incident fixes: vendor accuracy claims are presented as roughly 82%, leaving meaningful risk during outages.

Tools not to prioritize learning

• Ingress NGINX — Retired; learn how to migrate to Gateway API instead.
• CDK for Terraform — Described as deprecated and no longer maintained. The presenter clarifies that AWS CDK is a different product.
• Grafana OnCall (open source) — Archived.
• Opsgenie — Sales ended and shutdown is underway, according to the presenter.
• tfsec — Not described as dead, but the presenter says not to start new work on it because its engineering effort moved to Trivy.
• Chef and Puppet — The presenter recommends learning Ansible instead.

Main speaker and sources

The main speaker is the video’s presenter, who gives a first-person analysis and recommendations. Sources mentioned include Google’s DORA report, an Oxylabs analysis of US job postings, CNCF surveys and project information, Kubernetes project announcements, and roadmap.sh. The presenter says supporting source links are provided in the video description.