yaooqinn commented on issue #786:
URL: 
https://github.com/apache/incubator-kyuubi/issues/786#issuecomment-878858464


   For engine, we'd call it service discovery, not load balancing. The engine 
discovery logic is better to bind with client requests, a.k.a, JDBC 
connections, as an engine is stateful with cache tables, udf, etc. We already 
have implemented 3 share levels, CONNECTION/USER/SERVER for different kinds of 
usage scenarios.
   - For CONNECTION mode, the engine and all states are within a JDBC 
connection.
   - For USER mode, the engine and all states might/could share across all JDBC 
connections of a user with a `subDomain` if any.
   
   For `The last engine is acquired by default`, it's a potential bug since 
`1.2.0-rc`s as a distributed lock for engine creation. When an engine is 
created but killed forcefully and externally, the service uri of this engine 
will not get cleared. And another thread might get a malformed engine after it 
gains the lock. For this part, I think we need to introduce a retry and 
recreation of engine
   


-- 
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