keith-turner commented on PR #6217:
URL: https://github.com/apache/accumulo/pull/6217#issuecomment-4158639897

   This change is very close to being done and ready for review.  The current 
problem that I am running into is dealing with how the compaction coordinator 
deals with deleting tmp files from failed compactions.  The current process is 
event driven and directly responds to failed compaction notifications.  This 
will not work well when these notification are now happening across all manager 
processes.   Also w/ the current code if the process dies before processing the 
notification then the tmp file may never be cleaned up.
   
   It may be better to create a new periodic process that  compares compaction 
tmp files in dfs to running compactions in the metadata and fate table.  
However, concerned about this scanning all of DFS periodically.  Was wondering 
about putting tmp files in a single dir for each table on each volume so they 
are easier to find.   This could be a periodic task the Accumulo GC runs.
   
   Probably need to address this in a separate PR before this can be completed.


-- 
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