capistrant commented on PR #12599:
URL: https://github.com/apache/druid/pull/12599#issuecomment-1156653270

   > #12404 (currently-open PR) also contemplates doing schema changes in the 
metadata store. @capistrant I see you've commented in there too. Will the 
feature in this PR be able to use that other PR's mechanism? (Once it's merged?)
   
   My plan is to use as much of this PRs process as is feasible. Our approaches 
are nearly identical for the actual alteration of the table on startup of the 
controlling service (overlord for that one, coordinator for mine). The 
difference really emerges in the migration process. The linked PR is more 
complicated than mine because it is replacing existing always-on functionality. 
That requires them to be deliberate about how they transition from querying sql 
the old way to the new way. They background a thread on startup that 
incrementally brings the table up to spec by populating rows. I could certainly 
implement a similar approach instead of allowing the operator to optionally 
make their existing unused segments compatible with the buffer period by using 
a druid cli tool or an adhoc sql statement. 
   
   Long story short, I don't think there is anything 12404 would/should take 
from my implementation. However, there is certainly shared code that I can take 
from them once they are merged. I will contemplate the idea of a background 
thread that brings the table up to spec, it would really be pretty straight 
forward. If there are no unused rows with null `last_used` exit immediately. 
Otherwise use 12404 incremental batching process to update with current date 
until there are no null values in the column. That would eliminate 
overhead/confusion for our operators


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to