get allocation ratio from n-c-c relation

[Laurent Sesquès]

nova-cloud-controller has allocation ratio about the cloud, so p-o-e should get them from there instead of manually setting options.

[Mihaela Andronic (mihaela)]

Grafana does not correctly display `ram-allocation-ratio` on the dashboard because it uses values from prometheus-openstack-export which are set statically
https://git.launchpad.net/~prometheus-charmers/prometheus-openstack-exporter-charm/+git/charm-prometheus-openstack-exporter/tree/config.yaml

At the moment the following actions have to be carried out when changes are done, or default doesn't match:
AP:
1. Audit the clouds for the 3x values form
$ juju config nova-cloud-controller cpu-allocation-ratio
$ juju config nova-cloud-controller ram-allocation-ratio
$ juju config nova-cloud-controller disk-allocation-ratio

2: Use the information obtained from the commands above to replace the defaults from prometheus-openstack-exporter with the correct values:
$ juju config prometheus-openstack-exporter cpu-allocation-ratio
$ juju config prometheus-openstack-exporter ram-allocation-ratio
$ juju config prometheus-openstack-exporter disk-allocation-ratio

For customer using Grafana as reference, the defaults can be misleading.

[Adam Dyess (addyess)]

Please ignore those statically config multipliers because they are still wrong.
They must come from the placement api PER hypervisor (resource allocation) - see LP1938078

[Drew Freiberger (afreiberger)]

Hopefully, we can use relations to nova-compute when this spec (https://review.opendev.org/c/openstack/charm-specs/+/806997) is implemented and we can convert nova-compute apps into dense vs dedicated nodes.

While it's agreed that placement API provides the most updated allocation ratio info, the dashboard cannot possibly be set to refresh each time an operator changes the config via the API. Ultimately, when 21.10 charms land, nova-compute app will have cpu-allocation-ratio and ram-allocation-ratio that can be provided over a relation.

I believe the proper work here is to group "schedulable instances" based on nova-compute applications.