Add float_service label to health-check probes

roles/float-infra-prometheus/templates/prometheus.yml.j2

@@ -135,6 +135,7 @@ scrape_configs:
           probe: health
           probeset: health
           prober_float_service: prometheus
+          float_service: "{{ service_name }}"
 {% endfor %}
 {% endfor %}
 

roles/float-infra-prometheus/templates/rules/rules_base.conf.yml

@@ -8,20 +8,22 @@ groups:
       - record: job:up:ratio
         expr: job:up:sum / job:up:count
 
-      # Sum prober metrics over the probers (hosts), producing
-      # an aggregation by target.
+      # Sum prober metrics over the probers (hosts), producing an
+      # aggregation by target. The following rules use 'without' in
+      # order to preserve additional probe_success labels that might
+      # be present.
       - record: target:probe_success:count
-        expr: count(probe_success) by (probe,probeset,zone,host,prober_float_service)
+        expr: count(probe_success) without (job,instance,prober_host)
       - record: target:probe_success:sum
-        expr: sum(probe_success) by (probe,probeset,zone,host,prober_float_service)
+        expr: sum(probe_success) without (job,instance,prober_host)
       - record: target:probe_success:ratio
         expr: target:probe_success:sum / target:probe_success:count
 
       # Sum prober metrics over targets, aggregating by probe.
       - record: probe:probe_success:count
-        expr: count(probe_success) by (probe,probeset,prober_float_service,zone)
+        expr: count(probe_success) without (job,instance,prober_host,host)
       - record: probe:probe_success:sum
-        expr: sum(probe_success) by (probe,probeset,prober_float_service,zone)
+        expr: sum(probe_success) without (job,instance,prober_host,host)
       - record: probe:probe_success:ratio
         expr: probe:probe_success:sum / probe:probe_success:count