Prometheus should keep metrics as long as Loki keeps logs

... which is both governed by our retention policy.

nixos/modules/monitoring/server/prometheus.nix

@@ -13,6 +13,7 @@ let
     regex = "^(.*)(?:\\.monitoringvpn):\\d+$";
     target_label = "instance";
   };
+  logRetention = toString(config.services.private-storage.monitoring.policy.logRetentionSeconds) + "s";
 
 in {
   options.services.private-storage.monitoring.prometheus = {
@@ -44,6 +45,7 @@ in {
     services.prometheus = {
       enable = true;
       # port = 9090; # Option only in recent (20.09?) nixpkgs, 9090 default
+      retentionTime = logRetention;
       scrapeConfigs = [
         {
           job_name = "node-exporters";

[Jean-Paul Calderone]

This is fine. Just a note (which you probably already know) that you can also do:

"${config.services.private-storage.monitoring.policy.logRetentionSeconds}s"

which maybe is more pleasurable to look upon, sometimes.

[Florian Sesser]

Does not work I am afraid: error: cannot coerce an integer to a string