Change the default thread pool size

In testing ironic it w/ mod-wsgi I've found that ironic does
not pool connections to the messaging backend. This causes
connection and intermitent issues within the ironic service.
To mitigate these issues I've changed the ironic_wsgi_threads
variable from using vcpus*2 to using vcpus//4 which will
instruct the ironic process to only use 25% of the available
vcpus rounded to the nearest whole number. A ternary operation
was added to the default value to ensure the computed value is
greater than or equal to 1.

Error Example:
    XXX_ironic_conductor_container-XXX ironic-conductor:
    2016-04-07 14:32:32.016 555 ERROR oslo.messaging._drivers.impl_rabbit
    [-] AMQP server on 172.19.102.219:5671 is unreachable:
    Basic.cancel: (0) 1. Trying again in 1 seconds.

With a reduced thread count this error was resolved. My hosts have
48vCPUs resulting in 96 threads being used in the process.

Change-Id: I0390f0d3f5d90f851f5373962c4b924d3ade72ff
Signed-off-by: Kevin Carter <kevin.carter@rackspace.com>

defaults/main.yml

@@ -154,7 +154,8 @@ ironic_service_user_name: "ironic"
 
 # Apache settings
 ironic_wsgi_threads: 1
-ironic_wsgi_processes: "{{ ansible_processor_vcpus | default (1) * 2 }}"
+_ironic_wsgi_processes: "{{ ansible_processor_vcpus | default(1) // 4 }}"
+ironic_wsgi_processes: "{{ (_ironic_wsgi_processes | int > 1) | ternary(_ironic_wsgi_processes, 1) }}"
 
 ### OpenStack Services to integrate with
 

templates/ironic.conf.j2

@@ -1893,7 +1893,7 @@ rabbit_virtual_host={{ ironic_rabbitmq_vhost }}
 
 # Maximum number of connections to keep queued. (integer
 # value)
-#pool_max_size=10
+pool_max_size = {{ ironic_wsgi_processes }}
 
 # Maximum number of connections to create above
 # `pool_max_size`. (integer value)