belliottsmith commented on code in PR #3967:
URL: https://github.com/apache/cassandra/pull/3967#discussion_r1993282230
##########
src/java/org/apache/cassandra/schema/SchemaKeyspace.java:
##########
@@ -140,6 +140,7 @@ private SchemaKeyspace()
+ "clustering_order text,"
+ "column_name_bytes blob,"
+ "kind text,"
+ + "unique_id int,"
Review Comment:
If the listener is deterministically executed by CMS then yes, I think that
is probably sufficient. If the `TableMetadata` is constructed on recipient
replicas separately (and they would be expected to allocate these ids),
however, that might be an issue as they might skip some epochs?
I think basically we just always need CMS to allocate the uids, as even in a
live cluster post transformation a recipient might have missed some intervening
epochs where uids are issued, and could allocate the uids in a different order.
If CMS is already the one applying the transformation we're fine, and the
listener approach seems sufficient (though, my familiarity with TCM is still
low so I could be missing something)
--
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]
