Skip to content

Latest commit

 

History

History
49 lines (31 loc) · 3.03 KB

aks-memory-saturation-after-upgrade.md

File metadata and controls

49 lines (31 loc) · 3.03 KB
title description ms.date editor ms.reviewer ms.service ms.custom
Memory saturation occurs after upgrade to Kubernetes 1.25
Resolve pod failures caused by memory saturation and out-of-memory errors after you upgrade an Azure Kubernetes Service (AKS) cluster to Kubernetes 1.25.x.
06/14/2023
v-jsitser
aritraghosh, cssakscic, v-leedennis
azure-kubernetes-service
sap:Create, Upgrade, Scale and Delete operations (cluster or nodepool)

Memory saturation occurs in pods after cluster upgrade to Kubernetes 1.25

This article discusses how to fix pods that stop working because of memory saturation or out-of-memory (OOM) errors that occur after you upgrade a Microsoft Azure Kubernetes Service (AKS) cluster to Kubernetes 1.25.x.

Symptoms

One or more of the following issues occur:

  • Memory pressure on nodes

  • Increased memory usage for apps when compared to their memory usage before the upgrade

  • CPU throttling on nodes

  • Pod failure because of OOM errors

Performance degradation can occur in apps that run in the following environments:

Note

This list of environments in which performance degradation can occur isn't a comprehensive list. There might be other environments that experience memory saturation or OOM issues.

Solution

Beginning in the release of Kubernetes 1.25, the cgroup version 2 API has reached general availability (GA). AKS now uses Ubuntu Linux version 22.04. By default, version 22.04 uses cgroup version 2 API. To make sure the cgroup version 2 API is available for use in other environments to prevent the memory saturation issue, follow this guidance:

In addition, to enable pods to use more resources, increase their memory requests and limits.

[!INCLUDE Third-party disclaimer]

[!INCLUDE Third-party disclaimer]

[!INCLUDE Azure Help Support]