ajothomas commented on PR #1708:
URL: https://github.com/apache/samza/pull/1708#issuecomment-2486952101

   > Now I think about it, I have a question about whether this can work for 
Samza incremental checkpoint. So assume the first checkpoint failed during the 
state upload, we suppress the exception and let it continue. Then the next 
checkpoint will invoke rocksDb.checkpoint() which will do a delta of the 
previous checkpoint. If the second checkpoint succeeds, but the first 
checkpoint actually fails, seems the states needed will not be in ambry anymore.
   > 
   > We need to have a in-depth discussion on this. Thanks.
   
   Hey @xinyuiscool This is a great point and something I missed in the 
original review. As far as I know, rocksdb checkpoints are not incremental 
snapshots but point-in-time complete snapshot so we should be okay even if the 
previous commit failed as the rocksdb checkpoint would have hard links to all 
ssts. 
https://rocksdb.org/blog/2015/11/10/use-checkpoints-for-efficient-snapshots.html


-- 
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