Example project that will serve files from a cloud storage bucket but caching them in groupcache. Imagine this being part of an image handling system that also did image processing - you could have a cache layer for source images to avoid hitting GCS each time and another cache layer for resized images to avoid re-processing them on each request, and each source-load and image-resize would only happen on a single node, with other nodes waiting for completion.

The example is simple and just dumps the bytes read from GCS to the response, with caching on the way. A more 'complete' system might use protobuf to store content type and other file information as a part of the cached item.

This example is based on deployment to Docker Optimized GCE instances. You should change any references to the project, zones and buckets to match your project.

Install and setup:

• Docker
• gcloud SDK
• govendor

Run govendor install to install the vendored dependencies required to build the docker container.

These are examples only to fire up some instances:

Set gcloud project:

gcloud config set project my-project

Build docker image and push to gcloud registry:

docker build -t gce-cache-cluster . docker tag gce-cache-cluster gcr.io/captain-codeman/gce-cache-cluster docker push gcr.io/captain-codeman/gce-cache-cluster

Create a GCE instance using this image:

gcloud compute instances create gce-cache-cluster-1 \
    --image-family cos-stable \
    --image-project cos-cloud \
    --zone us-central1-b \
    --machine-type f1-micro \
    --tags http-server \
    --metadata service=test \
    --metadata-from-file user-data=cloud-init.txt

See cloud-init.txt for service and networking settings.