Comment 0 for bug 1914303

* A few references on the issue:

  * https://www.robustperception.io/high-availability-prometheus-alerting-and-notification#more-3014
  * https://image.slidesharecdn.com/promcon-180811113528/95/taking-advantage-of-prometheus-relabeling-37-638.jpg?cb=1533987350

* The issue is that the prometheus.yml file is jinja2 templated in a weird way [0].
  The "monitor" label is based on the service name, which means that we cannot really aggregate this without the admin indicating what he wants to do with this label.
  Plus there is the question of the other external labels that can be set in extra.

  I don't see right away a good option to "fix" this. This software is just too complex to be smarter than "put the raw configuration in a juju config" path IMHO.

  The external label are usually used to avoid duplicates when a storage aggregate the metrics from several datacenters.
  I don't see the value of having an automatic external label that is set automatically based on the name of the service.

  I would suggest the following solution:

  * remove the "monitor" external label et let the juju config item external-labels be
  * add another alert-relabel-config juju items, that would include a string, getting the yaml that would be used to configure the section