NodePool queries to the OpenStack are represented as task, enqueued and processed one by one at a given rate (eg every 10 seconds).
I have noticed the task ListFloatingIPsTask which is to grab the list of IPs (public?), they are cached internally for 5 seconds and are a major source of API request.
However, we have no floating IP allocated:
$ nova absolute-limits|grep FloatingIps | FloatingIps | 0 | 0 |
Internally Nodepool will retrieve them whenever the Nova has the extension os-floating-ips.
We should most probably hack our Nodepool to stop issuing those requests.