Details
-
New Feature
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Won't Do
-
None
-
None
Description
Looks like readwritesplit router assumes that Galera nodes are all always in sync. In reality they may have write sets pending to be applied and wsrep_local_recv_queue may be used as a measure of current "lagging" for Galera node.
With this we may be able to apply slave_selection_criteria = LEAST_BEHIND_MASTER and route to the Galera node that is more close to "master" or less loaded.