Comment 5 for bug 1168291

Ovais Tariq (ovais-tariq) wrote :

Hi Raghu,

For events that come in via replication, its perfectly valid to use the auto_increment* variables as set on the master. I do not see why we shouldn't be doing that, this is how its managed in async replication, the master also injects its auto_increment* variables and the slave SQL_threads sets those for its session, I would think of that as the correct way in the context of PXC cluster as well. Any SET SESSION .... in the binary log should be honoured and should not be overridden by the global variable if the variable is allowed to be changed dynamically because that's how variable scoping works in MySQL :)