Configure Prometheus federation to aggregate metrics from multiple Prometheus servers into a global view

Steps

1. On each source Prometheus server, the built-in /federate endpoint is always available; it exposes the current value of selected time series based on match[] URL parameters.
2. On the global/aggregating Prometheus, add a scrape_configs job with honor_labels: true and metrics_path: '/federate', targeting the source servers' host:port as static_configs targets.
3. Add one or more params: match[] entries to select which series to pull, e.g. match[]: ['{job="prometheus"}'] or match[] on specific metric name prefixes; multiple match[] values are unioned.
4. Set honor_labels: true so that labels already present on the scraped (federated) series take precedence over the global Prometheus's own server-side labels, preserving the original instance/job identity.
5. Reload the global Prometheus config and confirm via its targets page that the federation scrape job is healthy and the federated series appear in queries.
6. Prefer federating aggregated/recording-rule metrics rather than raw high-cardinality series, since federation is a pull-based full scrape of the matched series on every scrape interval.