[HDDS-2680] Fix updating lastAppliedIndex in OzoneManagerStateMachine

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Labels: pull-request-available
• Assignee: Bharat Viswanadham
• Reporter: Bharat Viswanadham
• Created: 05/Dec/19 22:32
• Updated: 01/Mar/24 19:08
• Resolved: 13/Dec/19 15:50

Description

This Jira is to fix a bug that has been caused by HDDS-2637.

Now we have applyTransactionMap where we put all entries of ratisTransactionIndex before complete.

 

And when updating lastAppliedIndex, now there is a chance that we updateLastAppliedIndex before DoubleBuffer flush has committed transactions to DB.

 

 

Let's take 1-10 are apply transaction entries which are not flushed, 11th is indexUpdate a metadata/conf entry transaction, now when notifyIndexUpdate is called with 11, we updateLastAppliedIndex to 11. (Which we should not do) This Jira is to fix this issue.

Issue Links

• causes: HDDS-9342 OM restart failed due to transactionLogIndex smaller than current updateID
• is fixed by: HDDS-9876 OzoneManagerStateMachine should add response to OzoneManagerDoubleBuffer for every write request
• is related to: HDDS-10019 Simplify updateLastAppliedIndex in OzoneManagerStateMachine
• links to: GitHub Pull Request #316