Uploaded May 2026 | Updated September 2026, 1 day ago
In this AMA livestream, Viktor and Scott tackle a wide range of DevOps and Kubernetes questions from the community. Topics include replacing a complex AWS stack (EKS, EFS, VPC peering, ECR, S3) with lighter alternatives, bootstrapping a self-managing EKS cluster using Crossplane and Minikube, setting up high-availability Kubernetes on VPS or home lab environments using tools like K3s and Cluster API, and the merits of Crossplane versus Terraform for infrastructure management. The duo also weigh in on observability tooling, comparing Prometheus, Victoria Metrics, Loki, and the OpenTelemetry ecosystem, while sharing thoughts on getting traction for a new open-source OTLP logging backend.
The conversation also covers practical advice on container resource requests and limits (and why VPA is strongly discouraged), guardrails for AI agents in Kubernetes environments using tools like Claw Patrol, the differences between upstream Crossplane and Upbound's paid tiers, and why ECS is considered unnecessarily complex compared to Kubernetes. Viktor and Scott also offer candid career advice for those transitioning into DevOps, including tips on interview confidence, honest CV presentation, and how to make the most of downtime at work by continuously learning and experimenting with new tools.
▬▬▬▬▬▬ ⏱ Timecodes ⏱ ▬▬▬▬▬▬
00:00 Intro (skip to first question)
06:04 How to replace EKS/EFS/VPC/ECR/S3 stack without rewriting microservices
08:17 Setting up self-managing EKS cluster using Crossplane and Minikube
11:04 Stakeholders want lighter AWS alternatives to avoid vendor lock-in
13:15 Using cert-manager with Keycloak and SAML vs OIDC
14:01 How to upgrade Kubernetes version on self-managing EKS cluster
15:02 Running highly available Kubernetes on cheap VPS instances
17:43 Building new OTLP log backend - how to gain community traction
23:19 Cheapest and most efficient way to run Kubernetes in production
26:57 Prometheus vs Victoria Metrics for monitoring
28:46 What do you gain moving from upstream Crossplane to UXP or paid Upbound
30:15 How to add foolproof guardrails for AI agents in Kubernetes
35:42 Can OpenTelemetry replace New Relic agent with Grafana stack
39:43 8 months Kubernetes experience, weak on monitoring, low interview confidence
44:12 Can you replace AWS stack with container-as-a-service like Cloud Run
45:27 What is the best way to learn DevOps and key skills needed
46:19 Crossplane configurations vs Terraform for platform provisioning
48:09 Prometheus agent in cluster vs Victoria Metrics outside cluster
49:35 30 services on ECS Fargate with bad performance - should we use EKS
54:11 Terraform vs Crossplane for managing infrastructure
57:05 How to donate a Crossplane provider to crossplane-contrib
58:13 Container resource requests and limits best practices
01:00:22 DevOps workday - what to do when you finish tasks early
01:11:11 Why is VPA horrible and what to use instead
01:13:44 Why should teams ditch ECS for Kubernetes
In this AMA livestream, Viktor and Scott tackle a wide range of DevOps and Kubernetes questions from the community. Topics include replacing a complex AWS stack (EKS, EFS, VPC peering, ECR, S3) with lighter alternatives, bootstrapping a self-managing EKS cluster using Crossplane and Minikube, setting up high-availability Kubernetes on VPS or home lab environments using tools like K3s and Cluster API, and the merits of Crossplane versus Terraform for infrastructure management. The duo also weigh in on observability tooling, comparing Prometheus, Victoria Metrics, Loki, and the OpenTelemetry ecosystem, while sharing thoughts on getting traction for a new open-source OTLP logging backend.
The conversation also covers practical advice on container resource requests and limits (and why VPA is strongly discouraged), guardrails for AI agents in Kubernetes environments using tools like Claw Patrol, the differences between upstream Crossplane and Upbound's paid tiers, and why ECS is considered unnecessarily complex compared to Kubernetes. Viktor and Scott also offer candid career advice for those transitioning into DevOps, including tips on interview confidence, honest CV presentation, and how to make the most of downtime at work by continuously learning and experimenting with new tools.
▬▬▬▬▬▬ ⏱ Timecodes ⏱ ▬▬▬▬▬▬
00:00 Intro (skip to first question)
06:04 How to replace EKS/EFS/VPC/ECR/S3 stack without rewriting microservices
08:17 Setting up self-managing EKS cluster using Crossplane and Minikube
11:04 Stakeholders want lighter AWS alternatives to avoid vendor lock-in
13:15 Using cert-manager with Keycloak and SAML vs OIDC
14:01 How to upgrade Kubernetes version on self-managing EKS cluster
15:02 Running highly available Kubernetes on cheap VPS instances
17:43 Building new OTLP log backend - how to gain community traction
23:19 Cheapest and most efficient way to run Kubernetes in production
26:57 Prometheus vs Victoria Metrics for monitoring
28:46 What do you gain moving from upstream Crossplane to UXP or paid Upbound
30:15 How to add foolproof guardrails for AI agents in Kubernetes
35:42 Can OpenTelemetry replace New Relic agent with Grafana stack
39:43 8 months Kubernetes experience, weak on monitoring, low interview confidence
44:12 Can you replace AWS stack with container-as-a-service like Cloud Run
45:27 What is the best way to learn DevOps and key skills needed
46:19 Crossplane configurations vs Terraform for platform provisioning
48:09 Prometheus agent in cluster vs Victoria Metrics outside cluster
49:35 30 services on ECS Fargate with bad performance - should we use EKS
54:11 Terraform vs Crossplane for managing infrastructure
57:05 How to donate a Crossplane provider to crossplane-contrib
58:13 Container resource requests and limits best practices
01:00:22 DevOps workday - what to do when you finish tasks early
01:11:11 Why is VPA horrible and what to use instead
01:13:44 Why should teams ditch ECS for Kubernetes
