1

We just started using Anthos as a service mesh for our Autopilot GKE cluster. The pods have previously requested 0.25vCPU with 512 Mi of memory.

However, after we enabled Anthos, the injected Istio sidecar container is requesting 512Mi of memory as well as 0.5vCPU; thus the sidecar requests twice the CPU our application container is requesting.

Is there a way to override this configuration? Now the total requested CPU in a single pod is 0.75 vCPU and 1Gi of memory, which is wasting a lot of compute resources given that it mostly uses 0.1vCPU in total on average at any one time, and the memory usage is upwards of at most 0.2Gi

muffe
  • 47
  • 7
  • Modify the container resource manifest. However, I do not recommend micro-managing resources. 500m is the default. I only recommend increasing that number. https://istio.io/latest/docs/setup/additional-setup/customize-installation/ – John Hanley Sep 08 '22 at 07:31
  • I see. However, the resource requests are made by istio-proxy injected by Anthos automatically. Thus there is no way to modify the resource requests for that particular container. – muffe Sep 09 '22 at 07:37
  • Please check this [doc](https://cloud.google.com/service-mesh/docs/troubleshooting/troubleshoot-resources) and [doc2](https://cloud.google.com/service-mesh/docs/anthos-service-mesh-proxy-injection) – Abhijith Chitrapu Sep 09 '22 at 11:24

0 Answers0