Network unreachable error while pulling from ECR

[paul-civitas]

Describe the bug

When attempting to pull an image from ECR, we get an network unreachable error.

seanchambers@MacBook-Pro .aws % finch pull 830786003228.dkr.ecr.us-east-1.amazonaws.com/inspire:8bdbe4b4acbbfe00d2079da763756fe224d60974
830786003228.dkr.ecr.us-east-1.amazonaws.com/inspire:8bdbe4b4acbbfe00d2079da763756fe224d60974: resolved    |++++++++++++++++++++++++++++++++++++++| 
manifest-sha256:dbd97a589135fa967daf69c13435067711a4c61ded7c08bf7e46fb597399e330:       done      |++++++++++++++++++++++++++++++++++++++| 
config-sha256:e8f0fa39c99204e633a8d02511d44a3def8528c33d2d85536c364748630edaaf:        downloading  |--------------------------------------|  0.0 B/20.8 KiB 
elapsed: 1.5 s                                         total: 6.6 Ki (4.4 KiB/s)                    
FATA[0001] failed to copy: httpReadSeeker: failed open: failed to do request: Get "https://830786003228.dkr.ecr.us-east-1.amazonaws.com/v2/inspire/blobs/sha256:34606fd0a15d10779b9ea9595128ff911dd5fcd73f1108db4b07fddb34c79740": dial tcp 3.86.125.4:443: connect: network is unreachable 
FATA[0001] exit status 1

Steps to reproduce

I can not reproduce this issue on my computer, but two of my colleagues are having this issue.

Expected behavior

Image pulls successfully.

Screenshots or logs

Logs already attached.

Additional context

I'm on an m1 mac. My colleague is on an m2 mac. We're both using the arm64 version of finch, verified by running: file -b $(which finch)

I'm on Mac OS 14.0 they're on 13.4.

[sean-civitas]

finch-support-20231116120608.zip

attached support file

[paul-civitas]

I wonder if this is related: https://github.com/containerd/containerd/issues/7676

[vsiravar]

Thanks for trying finch. Is this observed only while pulling that particular image or is it a problem for any image that you try to pull from the registry?

[vsiravar]

I wonder if this is related: containerd/containerd#7676

Does your image also contain nondistributable layers. A link to the Dockerfile or the layer that is failing would help in the debugging effort.

[paul-civitas]

I checked and it doesn't have any non distributable layers.

I might be able to get you the image in question through AWS support if you're an amazon employee.