With my metricsinfra long-term end goal (allow any Cloud VPS project administrator to set arbitrary prometheus scrape targets and alerting rules for their project in a self-service fashion), defining Prometheus and Alertmanager configuration in metricsinfra hiera doesn't have all the features needed (namely the self-service part), so alternative solutions are required.

My solution for this is prometheus-configurator which in the long term will talk to some API/database and enable self-service configuration. This task is to get it set up on metricsinfra with both Prometheus and Alertmanager, with a static puppet-created config file instead of a database. It will temporarily increase setup complexity without much gain in itself, but makes other wanted functionality possible which is not possible with puppet-generated config.

This unfortunately closes the possibilities to share most Puppet code with production (see T266050#6565343), but I think that the possibilities doing it this way are worth the additional complexity.