[YARN-676] [Umbrella] Daemons crashing because of invalid state transitions

• Type: Task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Assignee: Unassigned
• Reporter: Vinod Kumar Vavilapalli
• Created: 13/May/13 03:45
• Updated: 26/Jul/13 11:49

Description

There are several tickets tracking invalid transitions which essentially crash the daemons - RM, NM or AM. This is tracking ticket.

We should try to fix as many of them as soon as possible.

Issue Links

• is related to: MAPREDUCE-5422 [Umbrella] Fix invalid state transitions in MRAppMaster

Sub-Tasks

• Handle ( or throw a proper error when receiving) status updates from application masters that have not registered: Closed
• RM crash with NPE on NODE_REMOVED event with FairScheduler: Closed
• Node Manager can not handle duplicate responses: Open
• Resource Manager throws InvalidStateTransitonException: Invalid event: CONTAINER_FINISHED at ALLOCATED for RMAppAttemptImpl: Closed
• Resource Manager throws InvalidStateTransitonException: Invalid event: APP_ACCEPTED at RUNNING for RMAppImpl: Resolved
• Node Manager throws org.apache.hadoop.yarn.state.InvalidStateTransitonException: Invalid event: RESOURCE_FAILED at DONE: Resolved
• InvalidStateTransitonException: Invalid event: INIT_CONTAINER at DONE for ContainerImpl in Node Manager: Resolved
• NodeManager has invalid state transition after error in resource localization: Closed
• RM crash with NPE on NODE_UPDATE: Closed
• Resource Manager throws InvalidStateTransitonException: Invalid event: CONTAINER_FINISHED at ALLOCATED: Resolved
• ResourceManager throws ArrayIndexOutOfBoundsException while handling CONTAINER_ALLOCATED for application attempt: Closed
• Cancelling ContainerLaunch#call at KILLING causes that the container cannot be completed: Closed
• ContainerImpl State Machine: Invalid event: CONTAINER_KILLED_ON_REQUEST at CONTAINER_CLEANEDUP_AFTER_KILL: Closed