Tutorial: Linux Memory Management and Containers - Gerlof Langeveld, AT Computing

Tutorial: Linux Memory Management and Containers - Gerlof Langeveld, AT Computing

When running containerized processes with Docker, Podman or Kubernetes, memory and swap limitations/reservations can be defined. These settings are effectuated through the cgroups (v2) memory controller that influences the memory behavior of such containerized process. In this tutorial, Gerlof covers the main principles of Linux memory management: demand paging, page scanning, swapping, swappiness, OOM killing, NUMA configurations, memory leakage, and memory categories like page cache, shared memory and tmpfs. After that he explains the relation with particular parameters that can be used when starting a container, like --memory, --memory-reservation, --memory-swappiness and --cpuset-mems (with Podman and Docker), as well as the relation with the fields pod.spec.containers.resources.limits.memory and pod.spec.containers.resources.requests.memory for Pods (with Kubernetes). During the presentation the implications of certain choices will be visualized with atop and other tools. Based on memory exercises, attendees can experience and simulate various memory situations on their own Linux systems. Everything needed for this will be in a Github repo that will be shared with the trainees during the tutorial. atop is part of most major Linux distributions or can be easily installed.