[YARN-4497] RM might fail to restart when recovering apps whose attempts are missing

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 2.9.0, 3.0.0-alpha1
• Hadoop Flags: Reviewed
• Assignee: Jun Gong
• Reporter: Jun Gong
• Created: 23/Dec/15 03:40
• Updated: 30/Aug/16 01:08
• Resolved: 22/Jan/16 15:00

Description

Find following problem when discussing in YARN-3480.

If RM fails to store some attempts in RMStateStore, there will be missing attempts in RMStateStore, for the case storing attempt1, attempt2 and attempt3, RM successfully stored attempt1 and attempt3, but failed to store attempt2. When RM restarts, in RMAppImpl#recover, we recover attempts one by one, for this case, we will recover attmept1, then attempt2. When recovering attempt2, we call ((RMAppAttemptImpl)this.currentAttempt).recover(state), it will first find its ApplicationAttemptStateData, but it could not find it, an error will come at assert attemptState != null(RMAppAttemptImpl#recover, line 880).

Issue Links

• is related to: YARN-4968 A couple of AM retry unit tests need to wait SchedulerApplicationAttempt stopped.
• relates to: YARN-3480 Recovery may get very slow with lots of services with lots of app-attempts