[YARN-1394] RM to inform AMs when a container completed due to NM going offline -planned or unplanned

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Component/s: graceful
• Assignee: Rohith Sharma K S
• Reporter: Steve Loughran
• Created: 08/Nov/13 09:27
• Updated: 15/Aug/16 18:05
• Resolved: 15/Aug/16 18:05

Description

YARN-914 proposes graceful decommission of an NM, and NMs already have the right to go offline.

If AMs could be told that a container completed from an NM option -offline vs decommission, the AM could use that in its future blacklisting and placement policy.

This matters in long-lived services which may like to place new instances where they were placed before, and track hosts failure rates

Issue Links

• is depended upon by: YARN-896 Roll up for long-lived services in YARN
• relates to: YARN-671 Add an interface on the RM to move NMs into a maintenance state