git clone git@github.com:mesosphere/playa-mesos.git
cd playa-mesos
vagrant up
vagrant ssh

sudo easy_install http://downloads.mesosphere.io/master/ubuntu/14.04/mesos-0.19.0_rc2-py2.7-linux-x86_64.egg

Example run:

mesos-http-proxy --help
mesos-http-proxy --service-url http://127.0.0.1:5000/

This proxy will periodically POST a payload like this to the /offer endpoint of your service:

{
  "attributes": [],
  "executor_ids": [
    "default"
  ],
  "framework_id": "20140715-151922-16842879-5050-6648-0011",
  "hostname": "10.141.141.10",
  "id": "20140715-151922-16842879-5050-6648-1942",
  "resources": {
    "cpus": 1.75,
    "disk": 34068.0,
    "mem": 936.0,
    "ports": [
      {
        "begin": 31000,
        "end": 32000
      }
    ]
  },
  "slave_id": "20140714-215541-16842879-5050-1243-0"
}

You will likely only care about the resources and attributes (if you set any in Mesos) keys. Your server should respond with a payload like:

[]

...to not run any jobs. Or:

[
  {
    "id": "my_job",
    "cmd": "pwd && echo hello, world",
    "resources": {
      "cpus": 0.25,
      "mem": 64
    }
  }
]

...to run my_job, claiming the stated amount of resources.

The proxy will also periodically POST status updates about job to the /status endpoint of your service:

{
  "id": "my_job",
  "state": "running"
}

The id will correspond to the id you returned in your /offer endpoint.

Possible states correspond directly to the Protocol Buffers version:

states = {
  0: "starting",
  1: "running",
  2: "finished", # terminal
  3: "failed", # terminal
  4: "killed", # terminal
  5: "lost", # terminal
  6: "staging",
}

Terminal means the job is no longer running or attempting to run.