[RATIS-1195] Log Entry Validation in appendEntry can fail after log purge.

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.0.0
• Component/s: server
• Assignee: Aravindan Vijayan
• Reporter: Aravindan Vijayan
• Created: 02/Dec/20 22:43
• Updated: 20/May/22 08:09
• Resolved: 08/Dec/20 08:23

Description

After a log purge is done to the last log index, RaftLog#getLastEntryTermIndex will return a null. In RaftLog#validateLogEntry, when the last term index is null, the expectation is that the new entry to be appended is lastSnapshotIndex + 1. However, the 'lastSnapshotIndex' in RaftLog is updated only through installSnapshot calls. Hence, the validation fails and all further appends will fail until a restart.

java.lang.IllegalStateException: Difference between entry index and RaftLog's latest snapshot index -1 is greater than 1 and in between log entries are not present, entry: term: 2
index: 326
stateMachineLogEntry {
  logData: "\b\v\022\000\032\023client-46A48A10F5C6\"\033\n\006hadoop\032\f192.168.48.7\"\003om1Za\n_\n\006hadoop\022\006hadoop\032\004vol3 \377\377\377\377\377\377\377\377\377\0012\017\b\001\022\006hadoop\032\001\200 \0002\016\b\002\022\005users\032\001\200 \0008\215\343\333\251\342.@\000H\000P\215\343\333\251\342.X\377\377\377\377\377\377\377\377\377\001"
  clientId: "\000\f\356MU\026<\331\234;\216\t\352\237\371\027"
  callId: 1
}

cc hanishakoneru / msingh

Issue Links

• is required by: HDDS-3698 Ozone Non-Rolling upgrades
• links to: GitHub Pull Request #322