I am interested in running HDFS using swarm. One of the requirements of HDFS is a Zookeeper ensemble. To operate a Zookeeper ensemble, each Zookeeper should have a unique ID, starting from 1 and ending in 255.

While it is possible to coordinate the ids using something like consul or etcd, it would be much simpler for swarm to pass in task information as environment variables into each task. For example:

• SWARM_TASK_ID
• SWARM_SERVICE_NAME
• ...

These variables can then be used by the containers to simplify orchestration.