User Avatar

OOM Kill

OomKill

The purpose of this scenario is to showcase resolution of an OOMKill without using Klaudia for guidance

Overview

When deploying applications to Kubernetes, it's common to encounter issues with memory allocation. These issues can arise from misconfigured resource requests and limits, leading to Out of Memory (OOM) errors. In this scenario, we use the Komodor platform to troubleshoot and resolve an OOMKill error.

Prerequisites

  • None

Problem

  • Navigate to the Services tab in the left menubar.
  • Select komodor-oomkilled
  • Notice this service goes from Healthy to Unhealthy in the timeline oomkill
  • Select Logs from the tab menu
  • Observe the amount of memory allocated to the pod
  • Notice the pod is trying to allocate 60M of memory

Resolution

  • Select the horizontal ellipsis on the top right of the screen, next to Restart
  • Select Configure Resources configure_resources
  • Observe the gap between the configured amount (40) and the allocated amount from the logs (60)
  • Adjust these values in the memory fields for the requests and limits columns to 60Mi
  • Select Update Resources update_resources
  • The change triggers a new deployment new_deployment
  • After a moment, the service returns to a healthy state healthy_service