0

I have some Windows Containers running on Kubernetes with the following memory requests and limits set in the deployment yaml file:

      containers:
        - name: $$DEPLOYMENT_ALIAS$$$$APP_NAME$$
          image: $$DOCKER_IMAGE_NAME$$:$$DOCKER_IMAGE_VERSION$$
          imagePullPolicy: "Always"
          resources:
            requests:
              memory: "2Gi"
              cpu: "1"
            limits:
              memory: "3Gi"
              cpu: "1.5"

However all of the pods get restarted when they reach 1GB consumption of memory.

I found another question stating that with docker, Windows Containers had memory constraints of 1GB when using Hyper-V Isolation Mode: Docker windows container memory limit

Is this also a thing for containers in kubernetes? Could there be any other reason why memory requests/limits are not applied for a Windows Container?

Observation: this project uses Azure Kubernetes Service (AKS) and I do not have access to the settings in this layer.

Eric Omine
  • 489
  • 4
  • 15
  • How's the overall shape of the cluster, ie. is it generally short on resources or you have a lot of headroom on each node? How do you measure the 1G consumption? – Rick Rackow Aug 11 '23 at 15:00
  • Yes, there's a lot of headroom in the cluster. Each node (2 pods per node) has 16GB of physical memory, which results in about 12GB of allocatable memory. So each pod should be able to handle 3GB of RAM. The 1GB memory consumption has been observed in the kubernetes dashboard. – Eric Omine Aug 11 '23 at 20:03

0 Answers0