Discussion with openstack team also raised the need to tune "worker-multiplier" on any api service. This charm config option sets the number of worker process per each service to 2x the number of CPUs by default. This will eat a lot of memory on a 12 cpu machine and create way too many workers on a smooshed deployment.
So far we have the percona config options to tune:
- max-connections
- innodb-buffer-pool-size
Discussion with openstack team also raised the need to tune "worker-multiplier" on any api service. This charm config option sets the number of worker process per each service to 2x the number of CPUs by default. This will eat a lot of memory on a 12 cpu machine and create way too many workers on a smooshed deployment.
So far we have the percona config options to tune: buffer- pool-size
- max-connections
- innodb-
And on all API services:
- worker-multiplier