Learn from production

• Kafka on Kubernetes: Performance Lessons for Any Disk-Heavy Data Service

  Jacob Amar

  This case study traces why Kafka started reading from disk after moving from EC2 to EKS, and lands on cgroup v2 reclaim pressure.

• When ETCD Crashes, Check Your Disks First: A Pod CrashLoopBack Debugging Story

  Anastassios Nanos

  This case study shows how etcd CrashLoopBack pods on a Karmada/k3s demo cluster turned out to be a ZFS I/O latency problem, and how four ZFS tuning settings fixed it — including the exact etcd Prometheus metrics to watch.

• Zero-Downtime EKS Migration: Production-Grade Kubernetes at Scale

  Krishnakanth E

  This case study explains how an ECS workload was migrated to EKS across two AWS regions with zero production downtime.

  It covers KEDA autoscaling, HashiCorp Vault, IRSA, disaster recovery, and a coordinated production cutover.

• We’re a 3-person tech team running production Kubernetes — So we built an AI SRE

  Leo Blondel

  This case study shows how a 3-person team built a Claude-based AI SRE that triages SigNoz alerts, checks Kubernetes, GitLab, logs, traces, and Slack, then ignores noise, escalates, or runs safe fixes.

More case studies →

Matching jobs

1. • DevOps Engineer with Mark43

   • Salary: $155K to $170K a year

   • Location: remote from

   • Tech stack: Kubernetes, Docker, Terraform

2. • DevOps Engineer with RobCo

   • Salary: US$70.74K to US$440K a year

   • Location: based in the office in Munich, DE

   • Tech stack: Kubernetes, AWS, Go, Python, Terraform, Datadog, Grafana, Prometheus

3. • Site Reliability Engineer with MyFitnessPal

   • Salary: $120K to $165K a year

   • Location: remote from

   • Tech stack: Kubernetes, AWS, Docker, Go, Python, Typescript, Terraform, GitHub Actions, Datadog

4. • AI Enterprise Technical Program Manager with Redhorse Corporation

   • Salary: $37 to $485.65K a year

   • Location: based in the office in Arlington, VA, USA

   • Tech stack: Kubernetes, AWS, Azure, Docker, Spark

5. • Commercial Account Executive with Vantage

   • Salary: $100K to $200K a year

   • Location: remote from

   • Tech stack: Kubernetes, AWS, CircleCI, Datadog

Discover more Kubernetes jobs on Kube Careers →

Build something

• Autoscalable GitLab runners on AWS EC2

  Timur Nizamutdinov

  This tutorial shows how to replace always-on EC2 runners with autoscaling GitLab runners using the AWS fleeting plugin and an Auto Scaling Group, so instances start per job and shut down when the queue is empty.

• Kubernetes KEDA Autoscaling: Scale Smarter, Not Harder

  Dharmendra Yadav

  This tutorial shows how KEDA scales workloads on queue depth instead of CPU, with a full RabbitMQ example: install, TriggerAuthentication, ScaledObject and a load test that proves the scaling works.

• Building a Distributed Bot Detection PoC Using Nginx Logs, awk, and Terraform on GCP

  Parag Shahade

  This tutorial shows how to build a simple bot-detection system from Nginx logs and use GCP controls to investigate and slow suspicious traffic.

• Testing Go Microservices with Real Production Traffic Using GoReplay

  Imesh Herath

  This tutorial shows how GoReplay replays real production traffic in Kubernetes so you can test a new Go service before sending users to it.

More tutorials →

Call for Papers closing soon

1. 2

   days

   Devopsdays Juiz de Fora

   The Call For Paper is open until 23 August 2026 at GMT-4. More info →

   • Location: Juiz de Fora, BR

   • In-person conference organized by Devopsdays.

   • The conference starts on the 17 October 2026.

   • Apply here

2. 3

   days

   ONE Summit Japan

   The Call For Paper is open until 24 August 2026 at GMT-4. More info →

   • Location: Tokyo, JP

   • In-person conference organized by CNCF.

   • The conference starts on the 11 December 2026.

   • Apply here

3. 3

   days

   Open Source Summit Japan 2026

   The Call For Paper is open until 24 August 2026 at GMT-4. More info →

   • Location: Tokyo, JP

   • In-person conference organized by Linux Foundation.

   • The conference starts on the 9 December 2026.

   • Apply here

4. 7

   days

   Monster Scale Summit 2027

   The Call For Paper is open until 28 August 2026 at GMT-4. More info →

   • This is a virtual event

   • Online conference organized by ScyllaDB.

   • The conference starts on the 11 March 2027.

   • Apply here

5. 9

   days

   Devopsdays Warsaw

   The Call For Paper is open until 30 August 2026 at GMT-4. More info →

   • Location: Warsaw, PL

   • In-person conference organized by Devopsdays.

   • The conference starts on the 23 November 2026.

   • Apply here

6. 9

   days

   Devopsdays Recife

   The Call For Paper is open until 30 August 2026 at GMT-4. More info →

   • Location: Recife, BR

   • In-person conference organized by Devopsdays.

   • The conference starts on the 12 December 2026.

   • Apply here

7. 9

   days

   Devopsdays Florianópolis

   The Call For Paper is open until 30 August 2026 at GMT-4. More info →

   • Location: Florianópolis, BR

   • In-person conference organized by Devopsdays.

   • The conference starts on the 24 October 2026.

   • Apply here

More articles

• The Illusion of CPU: Why HPA Fails for Pub/Sub Consumers and What to Do Instead

  Sujith K. Surendran

  This article explains why CPU is the wrong scaling signal for Pub/Sub consumers, since pods look healthy while the backlog grows, and argues for scaling on queue depth with KEDA instead.

• Go Fiber Benchmark Deep-Dive: Breaking 1 Million RPS on Apple Silicon

  Max Taylor

  This article explains how Go Fiber reached 1.16M RPS on Apple Silicon by running client and server benchmarks inside Docker Desktop's Linux VM to bypass macOS loopback TCP limits.

• Service Mesh Showdown: We Picked Istio. We’d Pick Linkerd Today.

  Moiz Ezzy

  An SRE compares Istio, Linkerd and Cilium after eighteen months of running Istio in production, on operational cost, performance overhead, security model and observability. The verdict: most teams should start with Linkerd, or no mesh at all.

• Inside Volcano Controllers: Gang Scheduling, State Machines, and Real Kubernetes Logs

  Talha Amjad

  This article opens up Volcano's controllers to show how gang scheduling avoids the deadlock where half a distributed job holds resources, following the PodGroup and Job state machines with real cluster logs.

Even more articles →