Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
1.1.5
-
None
Description
When running the benchmark suite using a connection pool (instead of establishing a new connection) it turns out that the connection time is significantly higher (depending on the number of connections in the pool) than performing a single connect.
This is caused by an unnecessary health check (sending MYSQL_PING) when iterating over the connections in the pool.
Instead iterating over the connections, we should keep a LRU list with unused connections, and an additional list for connection in use. When getting a connection from unused connections list, status should be checked via MYSQL_PING.
Attachments
Activity
Field | Original Value | New Value |
---|---|---|
Status | Open [ 1 ] | In Progress [ 3 ] |
Comment | [ [~diego dupin] Should we add an option in connector benchmarks for switching between connection pool and simple connection? Or should I add new tests? ] |
Resolution | Fixed [ 1 ] | |
Status | In Progress [ 3 ] | Closed [ 6 ] |
Same issue was reported on Stackoverflow