[MESOS-2067] Add HTTP API to the master for maintenance operations.

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.25.0
• Component/s: master
• Labels: mesosphere
• Target Version/s: 0.25.0
• Epic Link: Maintenance
• Sprint: Mesosphere Sprint 16, Mesosphere Sprint 17
• Story Points: 8
• Assignee: Joseph Wu
• Reporter: Benjamin Mahler
• Created: 11/Nov/14 02:19
• Updated: 25/Sep/15 22:46
• Resolved: 31/Aug/15 17:27

Description

Based on MESOS-1474, we'd like to provide an HTTP API on the master for the maintenance primitives in mesos.

For the MVP, we'll want something like this for manipulating the schedule:

/maintenance/schedule
  GET - returns the schedule, which will include the various maintenance windows.
  POST - create or update the schedule with a JSON blob (see below).

/maintenance/status
  GET - returns a list of machines and their maintenance mode.

/maintenance/start
  POST - Transition a set of machines from Draining into Deactivated mode.

/maintenance/stop
  POST - Transition a set of machines from Deactivated into Normal mode.

/maintenance/consensus <- (Not sure what the right name is.  matrix?  acceptance?)
  GET - Returns the latest info on which frameworks have accepted or declined the maintenance schedule.

(Note: The slashes in URLs might not be supported yet.)

A schedule might look like:

{
  "windows" : [
    {
      "machines" : [
          { "ip" : "192.168.0.1" },
          { "hostname" : "localhost" },
          ...
        ], 
      "unavailability" : {
        "start" : 12345, // Epoch seconds.
        "duration" : 1000 // Seconds.
      }
    },
    ...
  ]
}

There should be firewall settings such that only those with access to master can use these endpoints.

Issue Links

• blocks: MESOS-3038 Resource offers do not contain Unavailability, given a maintenance schedule
• blocks: MESOS-3265 Starting maintenance needs to deactivate agents and kill tasks.
• contains: MESOS-1476 Provide endpoints for deactivating / activating slaves.
• is blocked by: MESOS-2075 Add maintenance information to the replicated registry.
• relates to: MESOS-3069 Registry operations do not exist for manipulating maintanence schedules