Cached image is not able to get ports inside the Pod

[ggupta-98]

Hi,

I was able to create a snapshot of with prefetched image as described here, there is one issue I am facing here when the image is trying to run inside the pod, its not able to get port to run on or there is some networking error.

The application is supposed to start a flask server inside the pod on port 8001 but when trying to port-forward the service outside for testing its failing also when seeing the logs inside the application the server is not starting.

I tried the command

kubectl exec <pod-name> -n namespace -- ps aux

it gave the following output

root           1  1.3  3.3 48040744 2172680 ?    Dsl  16:04   0:08 /usr/bin/python3 /usr/local/bin/flask run --host=0.0.0.0 --port=8001

Can someone help me out here?

[yubingjiaocn]

Pre-fetched image doesn't change the image itself. Is the same image able to run on vanilla Bottlerocket OS?

[ggupta-98]

The image is running on normal AL2023 nvidia ami, when its pulling from ECR but not running the when doing with bottlerocket instance.
Should I try with just the bottlerocket instance without the snapshot?

[yubingjiaocn]

Yes, please launch a Bottlerocket instance using managed node group or Karpenter (if run in EKS), and try to run your workload. If the issue persists, there may be conflict between Bottlerocket OS's config with your own workload.

[ggupta-98]

Hi @yubingjiaocn without the caching the image is working fine in bottle rocket OS and I am able to invoke the application inside the pod. But something is happening when I am caching the image can you help me out here.

[ggupta-98]

Hi @yubingjiaocn can you help me out here

[yubingjiaocn]

Hi, could you please provide me your Dockerfile or image so I can test it in my environment? Thx.