yuqi1129 commented on issue #12020:
URL: https://github.com/apache/gravitino/issues/12020#issuecomment-5043032032

   Consistency clarification for this subtask: one shared Redis cache copy 
removes cross-node cache-copy divergence, but it does not by itself provide 
strong read-your-writes consistency relative to the metadata database.
   
   Before implementation, please define and test:
   
   - the write completion point and ordering of DB commit versus Redis 
fence/invalidation;
   - behavior when the DB succeeds but Redis times out or fails (request 
failure, fail-closed, or cache bypass);
   - a persistent version fence/tombstone that survives value deletion, 
hierarchical drop, rename, and value TTL, so an older in-flight load cannot 
refill the key;
   - Redis Cluster hash-slot constraints for every multi-key Lua operation;
   - concurrent reader/writer, delete-then-late-fill, Redis failure, and retry 
tests.
   
   #12149, #12150, and #12151 define the shared eligibility, fresh-read, and 
mutation-event contracts. If DB and Redis are not joined by a protocol that 
enforces the stated post-write guarantee, the issue should describe Redis as a 
shared cache rather than a strongly consistent database/cache pair.
   


-- 
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