klsince commented on issue #15846:
URL: https://github.com/apache/pinot/issues/15846#issuecomment-2949949500

   Good point about the need for enforcing seqid in name. Although subclasses 
could extend shouldReplaceOnComparisonTie() to check those, but the initial 
logic in the base class is better to be fixed anyway.
   
   While understanding why creation.meta files could be different even for 
committed segments, I see that the creation.meta file is not included when 
computing segment crc value. Otherwise different crc values would trigger the 
slow servers to sync up their segment replicas with the committed segments. 
Including creation.meta in crc computation could be costly, as the servers 
would pretty much always see different crc values and have to download segment 
to sync up replicas.
   
   Looking into the idea proposed in [the 
PR](https://github.com/apache/pinot/pull/16006), instead of using an explicit 
Map, I think we may try to check and fix the creation.meta file during segment 
commit process (for the slower replicas). Like where crc is checked, we can 
double check creation time there too. Then the slower replicas can init the 
SegmentMetadata object to have the right creation time, before finish the 
commit process with segment it's built locally. (Feel free to tag me in the PR 
to continue the discussions there)


-- 
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: commits-unsubscr...@pinot.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org
For additional commands, e-mail: commits-h...@pinot.apache.org

Reply via email to