[MESOS-887] Scheduler driver should use exited() to detect disconnection with Master.

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.13.0, 0.14.0, 0.14.1, 0.14.2, 0.15.0, 0.16.0
• Component/s: framework, master
• Labels: reliability twitter
• Epic Link: Master ↔ Scheduler Partitions
• Assignee: Unassigned
• Reporter: Benjamin Mahler
• Created: 19/Dec/13 02:46
• Updated: 02/Dec/16 22:17

Description

The Scheduler Driver already links with the master, but it does not use the built in exited() notification from libprocess to detect socket closure.

Of particular concern is that, if the socket breaks and subsequent messages are successfully sent on ephemeral sockets, then we don't re-register with the master. Messages may have been dropped.

Issue Links

• is cloned by: MESOS-1963 Slave should use exited() to detect disconnection with Master.
• is related to: MESOS-2352 Scheduler::disconnected() should be called when the single master fails
• is superceded by: MESOS-5180 Scheduler driver does not detect disconnection with master and reregister.
• relates to: MESOS-6676 Always re-link with scheduler during re-registration.