There are four examples in this folder.

This example will just start a command-based scheduler, which simply calls sleep 10. After execution of this command, the task will switch to TASK_FINISHED state.

To make the finished tasks restart again, we used the restartStates array definition within the task object. You can specify the list of task states which should trigger a restart of a task (by default, TASK_FINISHED doesn't trigger a restart of the task).

A simple Docker-based scheduler, which starts a simple Node.js webserver. It uses BRIDGE networking to allocate and map container to host ports.

A simple Docker-based scheduler, which starts a simple Node.js webserver. It uses HOST networking to allocate and use host ports.

A more advanced setup, which uses different task type with priorities. This scheduler start a Apache Flink cluster in HA mode (3 jobmanagers and 2 taskmanagers).

The only two things you should need to change to run this on your cluster are the following:

• masterUrl: Change this either to your leading master's ip address, or use leader.mesos in case you're using Mesos DNS.
• flink_recovery_zookeeper_quorum: This is the environment variable passed to the Docker containers to enable them to use HA via your existing ZooKeeper cluster (format is hostIp1:2181,hostIp2:2181,hostIpn:2181).