[MESOS-1962] Improve the partition tolerance between the Master and Schedulers

• Type: Epic
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: framework, master
• Epic Name: Master ↔ Scheduler Partitions
• Assignee: Unassigned
• Reporter: Benjamin Mahler
• Created: 22/Oct/14 01:05
• Updated: 02/Dec/16 22:16

Description

In a similar spirit to the master ↔ slave partition tolerance techniques, recently improved in MESOS-1529, the master and scheduler driver should better handle partitions.

In particular, heartbeats are needed to determine the liveness of the connection.

The goal here is to notify the scheduler when the connection is not healthy (via the disconnected callback), and to trigger a re-registration when either side of the links break.