[MESOS-3265] Starting maintenance needs to deactivate agents and kill tasks.

• Type: Task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.25.0
• Component/s: agent, master
• Labels: mesosphere
• Target Version/s: 0.25.0
• Epic Link: Maintenance
• Sprint: Mesosphere Sprint 17, Mesosphere Sprint 18
• Story Points: 8
• Assignee: Joris Van Remoortere
• Reporter: Joseph Wu
• Shepherd: Benjamin Hindman
• Created: 14/Aug/15 22:55
• Updated: 26/Nov/18 12:20
• Resolved: 26/Nov/18 12:20

Description

After using the /maintenance/start endpoint to begin maintenance on a machine, agents running on said machine should:

• Be deactivated such that no offers are sent from that agent. (Investigate if Master::deactivate(Slave*) can be used or modified for this purpose.)
• Kill all tasks still running on the agent (See MESOS-1475).
• Prevent other agents on that machine from registering or sending out offers. This will likely involve some modifications to Master::register and Master::reregister.

Issue Links

• blocks: MESOS-3266 Stopping/Completing maintenance needs to reactivate agents.
• is blocked by: MESOS-2067 Add HTTP API to the master for maintenance operations.
• Is contained by: MESOS-2076 Implement maintenance primitives in the Master.
• relates to: MESOS-2077 Ensure that TASK_LOSTs for a hard slave drain (SIGUSR1) include a Reason.