kadirozde commented on PR #5545:
URL: https://github.com/apache/hbase/pull/5545#issuecomment-1998519608

   @Apache9, I have moved historical file generation to StoreFileWriter. Based 
on a new config param called hbase.enable.historical.compaction.files (please 
suggest a better name if you like), StoreFileWriter uses a single file (just 
live file) or two file (live vs historical). This new config param replaces the 
previous one (hbase.hstore.defaultengine.enable.dualfilewriter). 
   
   I also changed the name of the new store file metadata key. Now it is called 
HISTORICAL. This is  set only when the file is historical. 
   
   Currently, we generate historical files only for compactions. However, we 
can enable it for flushes too. We can introduce another config param say 
base.enable.historical.flush.files to enable it for flushes. It will be a 
simple change. Please let me know if you like me to include this change too.
   
   Since historical file generated at the store file, it is available for all 
types of store engines and compactions by default. I have not disabled it for 
specific types of compactions yet. Either we can disable it for them or add 
tests to cover them. Let me know how you like to proceed.
   
   I also introduced a comprehensive data integrity checks with a new 
integration test class called TestStoreFileWriter. The test identified data 
integrity issues with version delete markers. I verified that these issues are 
not introduced by this PR and currently exist in hbase. I will file a jira for 
this. 
   
   After your feedback, I will update the design doc and PR. 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to