Evaluating Kubernetes Built-In Self-Healing Across Local and Cloud Environments Under Common Microservice Failure Scenarios

Coker, Oluseyi Iyetomide Simon (2025) Evaluating Kubernetes Built-In Self-Healing Across Local and Cloud Environments Under Common Microservice Failure Scenarios. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (771kB) | Preview

Abstract

Kubernetes has become a common platform for running microservices, mostly because it does a lot of the heavy lifting when things fail. One of the main reasons people trust it is its ability to fix basic problems on its own. When a pod crashes or becomes unhealthy, Kubernetes quietly replaces it and keeps the service running. Even though developers rely on this behaviour every day, there is surprisingly little work that looks closely at how well Kubernetes performs without any extra support from tools such as Istio, RAMSES or other resilience add-ons. This study tries to fill that gap by examining how Kubernetes copes with a few realistic microservice failures on its own. To do that, I used two setups. The first was a local Minikube cluster running through my own machine, and the second was a small K3s cluster on an EC2 instance. I deployed the same simple microservice on both and tested four scenarios: a clean baseline run, a pod crash, resource pressure using CPU and memory stress, and a network issue created by adding delay and packet loss. Each test was run under steady load so that I could measure things like recovery time, throughput and response latency. The results were fairly consistent across both environments. Kubernetes managed to recover from all four scenarios and kept the service available most of the time. The cloud setup recovered slightly faster, most likely because it had more stable resources to work with. When I compared these findings with results from studies that rely on external resilience tools, it became clear that Kubernetes does quite well on its own, although larger or more demanding systems would still benefit from additional layers of protection.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Siddig, Abubakr UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing
Divisions:	School of Computing > Master of Science in Cloud Computing
Depositing User:	Ciara O'Brien
Date Deposited:	31 Aug 2026 12:34
Last Modified:	31 Aug 2026 12:34
URI:	https://norma.ncirl.ie/id/eprint/9694

Actions (login required)

View Item