Closed gshiva closed 6 months ago
I recreated the cluster and it is working now. One difference is that I configured monitoring as soon as I created the cluster and did not wait for days to configure the monitoring. Another change I made is to give the Linux Node 8GB memory instead of the default 4GB. Not sure if those were the issue.
I will close this in a week if there is no response from the MS team.
Issues with pulling the images from aksiotdevacr.azurecr.io could indicate problems. Specially if the linux nodes resources are thin.
We believe it can be reproduced by allocating the default memory and storage space (4GB/10GB respectively) to the Linux node and then turning on Azure Arc monitoring. We can see that the disk space used is 18GB and memory used is 5.4GB.
Suggestions:
Once the resource constraints are removed, we dont see much of error messages for pulling images from aksiotdevacr.azurecr.io.
Describe the bug I enabled Azure Arc Monitoring integration via the Azure Portal. I see hundreds of pods in various error states. I recreated a pod definition and launched it to debug the issue. It is failing because it is unable to pull the image from aksiotdevacr.azurecr.io.
To Reproduce Steps to reproduce the behavior:
kubectl apply -f .\test-aks-pull.json
kubectl get events
Expected behavior The pod should be launched without any error.
Environment (please complete the following information):
Additional context I see the same error in multiple azure-arc pods.
pause
pod is just an example. The machine is behind a corp proxy. Not sure if that is an issue.