daniel.zuniga@gmail.com wrote:
Thanks for the guidance. Redeploying using MDB seems to have solved the performance problem.
I have to admit, it is a little unsettling to see near-zero CPU utilization vs 100% yet the same number of requests are being supported.
Your install of BerkeleyDB probably defaulted to using spinlock mutexes. Definitely wastes a ton of CPU, without getting anything useful done, particularly in a multi-core or multi-socket machine.
We ended up sticking to the RH/CentOS7 distribution of OpenLDAP but updated to whatever their latest supported release is.
Thank you!