This repo provides an infrastructure that reliably replicates coreos/bugs#2457. Or at least I think it is that issue, I've only seen it present as stalled docker pulls when using non ebs optimized images and an s3 backed docker registry.

How to Build

The below was tested with terraform v0.11.7

Building This Will Cost You Money!

## initialize
terraform init

## export keypair to use
export export TF_VAR_ssh_key=<keypair>

## build
terraform apply

## destroy
terraform destroy

Replicate Issue

First: Push an Image with Large Layers

ssh core@<staging_server_ip>

## using this because it is an image with some large layers
docker pull paulcichonski/docker-logstash-kafka:latest

docker tag paulcichonski/docker-logstash-kafka:latest <registry_ip>:5000/docker-logstash-kafka:latest

## if this fails you might need to restart the registry because timing issues, just ssh to registry and `sudo systemctl restart registry`
docker push <registry_ip>:5000/docker-logstash-kafka:latest

Second: Try to pull it

This will hang:

ssh core@<server_1745_6_0_broken_ip>

docker pull <registry_ip>:5000/docker-logstash-kafka:latest

This won't:

ssh core@<server_1745_5_0_working_ip>

docker pull <registry_ip>:5000/docker-logstash-kafka:latest