This question has probably been asked 100 times but here goes. We are having strange memory issues on our percona-galera cluster. Every so often we see chunks of memory being consumed. We have extremely high spec servers (32 cores 32GB RAM) so ample capacity. What I would like to do is getting a better idea of what is consuming the memory, it could be a bad query but nothing out of the ordinary is running.
We currently have the innodb buffer set to 4GB, we originally had this as 20GB but had the same issues.
There are a number of caches and buffers that get allocated. Normally the buffer_pool is the biggest item in memory, but if you have tweaked other settings, you could be causing the trouble. Would you like to share SHOW VARIABLES; and SHOW GLOBAL STATUS;. We might be able to spot something.