[KUDU-1469] Consensus can get "stuck" after a leader change when committed index is far behind replicated

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 0.8.0
• Fix Version/s: 0.9.1, 0.10.0
• Component/s: consensus
• Target Version/s: 1.0.0
• Code Review: http://gerrit.cloudera.org:8080/3228
• Assignee: Todd Lipcon
• Reporter: Todd Lipcon
• Created: 26/May/16 07:17
• Updated: 14/Jul/16 16:02
• Resolved: 22/Jun/16 23:43

Description

A YCSB stress test hit a bug whereby a tablet was "stuck" - operations were left in flight for hours/days and no progress could be made, despite a majority of replicas being alive.

Detailed analysis below.