The puppet module WSGI default values a basically useless and just asking for pain.
The WSGI threads behaves as cPhython threads, so they are suffering from the GIL related issues, and usually performer worse when you run more than 1 cpu intensive task at the same time.
The apache itself shields the workers from slow clients, unless you have very slow backed services (mysql) or very slow internal network, keystone will be cpu intensive.
Devstack also uses $subject style, so if you are using a different one just walking on less used path.
On a 8 CPU thread machine the WSGIDaemonProcess line needs to look like:
The puppet module WSGI default values a basically useless and just asking for pain.
The WSGI threads behaves as cPhython threads, so they are suffering from the GIL related issues, and usually performer worse when you run more than 1 cpu intensive task at the same time.
The apache itself shields the workers from slow clients, unless you have very slow backed services (mysql) or very slow internal network, keystone will be cpu intensive.
Devstack also uses $subject style, so if you are using a different one just walking on less used path.
On a 8 CPU thread machine the WSGIDaemonProcess line needs to look like:
WSGIDaemonProcess keystone_admin display- name=keystone- admin group=keystone processes=8 threads=1 user=keystone