Simple centralized cloud configuration server for Kubernetes in Go.

This configuration server exposes fast cached access to a slow backend.

• The server is meant to run as multiple replicas service in a Kubernetes cluster.

• groupcache is used as fast distributed cache. If multiple clients request a new configuration file, only a single replica fetches it from the backend and then makes the file available for all pending requests.

• The pod replicas automatically find each other by querying Kubernetes API for pods with a shared label app=<app-name>. For example, if a deployment is used to create the replicas, the shared label would be app=<deployment-name>.

• The server handles refresh notification events from the AMQP queue config-event-queue below. Whenever a refresh notification is received for an application, cache entries with that application configuration file are cleared, forcing their refresh from the backend.

exchangeName: springCloudBus
exchangeType: topic
queue:        config-event-queue

• The env var TTL can be used to enforce a TTL on cache entries. Example: TTL=300s. Default value is TTL=0, meaning no expiration set for cache entries.

git clone https://github.com/udhos/kubecloudconfigserver

cd kubecloudconfigserver

go mod tidy

go test -race ./...

export CGO_ENABLED=0

go install ./...

Proxying to an HTTP server:

export BACKEND=http://configserver:9000

kubeconfigserver

Reading from filesystem rooted at directory samples:

export BACKEND=dir:samples
export BACKEND_OPTIONS=flatten ;# flatten request path into a flat dir

kubeconfigserver

Query server:

curl localhost:8080/deploy.yml

Query groupcache directly:

curl localhost:5000/_groupcache/<group>/<key>

curl localhost:5000/_groupcache/configfiles/deploy.yml

Docker hub:

https://hub.docker.com/r/udhos/kubecloudconfigserver

Pull from docker hub:

docker pull udhos/kubecloudconfigserver:0.0.0

Build recipe:

./docker/build.sh