This is a variation on the classic "ELK" stack.
The fluentd subdirectory contains a Dockerfile to build a fluentd image embarking a simple configuration file, accepting log entries on port 24224 and storing them in Elasticsearch in a format that Kibana can use.
You can also use a pre-built image, jpetazzo/fluentd:v0.1
(e.g. if you want to deploy on a cluster and don't want to deploy
your own registry).
Once this fluentd container is running, and assuming you expose its port 24224/tcp somehow, you can send container logs to fluentd by using Docker's fluentd logging driver.
You can bring up the whole stack with the associated Compoes file. With Swarm mode, you can bring up the whole stack like this:
docker network create efk --driver overlay
docker service create --network efk \
--name elasticsearch elasticsearch:2
docker service create --network efk --publish 5601:5601 \
--name kibana kibana
docker service create --network efk --publish 24224:24224 \
--name fluentd jpetazzo/fluentd:v0.1And then, from any node on your cluster, you can send logs to fluentd like this:
docker run --log-driver fluentd --log-opt fluentd-address=localhost:24224 \
alpine echo ohai there