[SAMZA-2117] Container failures due to incorrect AM accounting

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.2
• Assignee: Jagadish
• Reporter: Jagadish
• Created: 28/Feb/19 18:49
• Updated: 11/Jun/19 22:12
• Resolved: 11/Jun/19 22:12

Description

The NMClient thread gets a callback from the NodeManager that the container has started even before the Container Allocator Thread puts it in the pending state.

The fix is to record state upfront before we issue requests - that way, the callback is guaranteed to see that state

Issue Links

• links to: GitHub Pull Request #958