We are experiencing what seems to be a memory leak after upgrading MySQL from 5.5 to 5.6. We have a cluster of MySQL instances, 2 per datacenter. In one of the datacenters, we upgraded 01 instance to 5.6 while 02 instance stayed on 5.5. Schema, data and my.cnf are identical on both instances. We use custom replication to those instances and not using MySQL replication slave. Application connections are load balanced across both instances. Aside from replication, queries are 100% reads.
Here is the graph of mysql process memory sizes of both instances, plus number of threads connected:
https://www.dropbox.com/s/il2swku5helrjl2/leak.png
Both instances were restarted around the same time to ilustrate the issue. Where MySQL 5.5 memory usage (green line) is staying stable, MySQL 5.6(red line) is constantly going up. At the same time, number of connections on both instances (blue and purple lines) are balanced pretty evenly.
Here is my.cnf:
http://pastebin.com/7itY4VfB
MySQL 5.6 version is 5.6.19
I am not sure where to proceed from here. There is obviously a leak. But how to identify which part of MySQL is leaking memory ? The only command I know that reports memory usage is "show engine innodb status" and InnoDB reports pretty much identical buffer pool size on both 5.5 and 5.6.
The only other note is that most of the queries are calls to stored procedures.
Limus
