naxos-os/modules/monitoring/perses.nix

{ config, pkgs, lib, ... }:

with lib;

let
  cfg = config.services.naxos.perses;
  
  # Declarative Perses Prometheus Datasource Manifest
  datasourceManifest = pkgs.writeText "perses-prom-datasource.json" (builtins.toJSON {
    kind = "Datasource";
    metadata = {
      name = "PrometheusDemo";
      project = "naxos";
    };
    spec = {
      default = true;
      plugin = {
        kind = "PrometheusDatasource";
        spec = {
          directUrl = "http://127.0.0.1:9090";
        };
      };
    };
  });

  # Declarative Perses ZFS & System Dashboard Manifest
  zfsDashboardManifest = pkgs.writeText "perses-zfs-dashboard.json" (builtins.toJSON {
    kind = "Dashboard";
    metadata = {
      name = "zfs-storage-health";
      project = "naxos";
    };
    spec = {
      duration = "1h";
      refreshInterval = "10s";
      display = {
        name = "ZFS Storage & System Health";
        description = "Native NaxOS analytics powered by Perses";
      };
      variables = [];
      panels = {
        cpuUsage = {
          kind = "Panel";
          spec = {
            display = { name = "CPU Utilization (%)"; };
            plugin = {
              kind = "TimeSeriesChart";
              spec = {
                queries = [{
                  kind = "TimeSeriesQuery";
                  spec = {
                    plugin = {
                      kind = "PrometheusTimeSeriesQuery";
                      spec = {
                        query = "100 - (avg by (instance) (rate(node_cpu_seconds_total{mode='idle'}[1m])) * 100)";
                      };
                    };
                  };
                }];
              };
            };
          };
        };
        memoryArc = {
          kind = "Panel";
          spec = {
            display = { name = "Memory & ARC Cache (Bytes)"; };
            plugin = {
              kind = "TimeSeriesChart";
              spec = {
                queries = [
                  {
                    kind = "TimeSeriesQuery";
                    spec = {
                      plugin = {
                        kind = "PrometheusTimeSeriesQuery";
                        spec = {
                          query = "node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes";
                        };
                      };
                    };
                  }
                ];
              };
            };
          };
        };
        diskIO = {
          kind = "Panel";
          spec = {
            display = { name = "ZFS Pool Read/Write Throughput"; };
            plugin = {
              kind = "TimeSeriesChart";
              spec = {
                queries = [{
                  kind = "TimeSeriesQuery";
                  spec = {
                    plugin = {
                      kind = "PrometheusTimeSeriesQuery";
                      spec = {
                        query = "rate(node_disk_read_bytes_total[1m]) + rate(node_disk_written_bytes_total[1m])";
                      };
                    };
                  };
                }];
              };
            };
          };
        };
      };
      layouts = [
        {
          kind = "Grid";
          spec = {
            items = [
              { x = 0; y = 0; width = 12; height = 6; content = { "$ref" = "#/spec/panels/cpuUsage"; }; }
              { x = 12; y = 0; width = 12; height = 6; content = { "$ref" = "#/spec/panels/memoryArc"; }; }
              { x = 0; y = 6; width = 24; height = 8; content = { "$ref" = "#/spec/panels/diskIO"; }; }
            ];
          };
        }
      ];
    };
  });

  persesConfigFile = pkgs.writeText "perses-config.yaml" ''
    database:
      file:
        folder: "/var/lib/perses"
        extension: "json"
    schemas:
      panels_path: "/var/lib/perses/schemas/panels"
      queries_path: "/var/lib/perses/schemas/queries"
      datasources_path: "/var/lib/perses/schemas/datasources"
      variables_path: "/var/lib/perses/schemas/variables"
    security:
      readonly: false
      enable_auth: false
  '';

in {
  options.services.naxos.perses = {
    enable = mkEnableOption "Perses Embedded Analytics Engine";

    port = mkOption {
      type = types.port;
      default = 8080;
      description = "Perses dashboard server port.";
    };
  };

  config = mkIf cfg.enable {
    # System user for Perses
    users.users.perses = {
      isSystemUser = true;
      group = "perses";
      home = "/var/lib/perses";
      createHome = true;
    };
    users.groups.perses = {};

    # Systemd service for Perses
    systemd.services.perses = {
      description = "Perses Native Observability and Dashboard Service";
      wantedBy = [ "multi-user.target" ];
      after = [ "network.target" "prometheus.service" ];
      serviceConfig = {
        User = "perses";
        Group = "perses";
        StateDirectory = "perses";
        WorkingDirectory = "/var/lib/perses";
        ExecStartPre = pkgs.writeShellScript "perses-provisioning" ''
          mkdir -p /var/lib/perses/projects/naxos/dashboards
          mkdir -p /var/lib/perses/projects/naxos/datasources
          cp -f ${datasourceManifest} /var/lib/perses/projects/naxos/datasources/PrometheusDemo.json
          cp -f ${zfsDashboardManifest} /var/lib/perses/projects/naxos/dashboards/zfs-storage-health.json
        '';
        ExecStart = "${pkgs.perses or pkgs.prometheus}/bin/perses --config=${persesConfigFile} --port=${toString cfg.port}";
        Restart = "always";
        RestartSec = "5s";
      };
    };

    networking.firewall.allowedTCPPorts = [ cfg.port ];
  };
}