Merge pull request #18770 from mayflower/prometheus-alertmanager

Prometheus alertmanager module

3 files changed, 144 insertions, 0 deletions

diff --git a/nixos/modules/module-list.nix b/nixos/modules/module-list.nix
index 6616238f50d..7064dcd968a 100644
--- a/nixos/modules/module-list.nix
+++ b/nixos/modules/module-list.nix
@@ -289,6 +289,7 @@
   ./services/monitoring/nagios.nix
   ./services/monitoring/prometheus/default.nix
   ./services/monitoring/prometheus/node-exporter.nix
+  ./services/monitoring/prometheus/alertmanager.nix
   ./services/monitoring/riemann.nix
   ./services/monitoring/riemann-dash.nix
   ./services/monitoring/riemann-tools.nix
diff --git a/nixos/modules/services/monitoring/prometheus/alertmanager.nix b/nixos/modules/services/monitoring/prometheus/alertmanager.nix
new file mode 100644
index 00000000000..a9c0ce4ed6c
--- /dev/null
+++ b/nixos/modules/services/monitoring/prometheus/alertmanager.nix
@@ -0,0 +1,116 @@
+{ config, pkgs, lib, ... }:
+
+with lib;
+
+let
+  cfg = config.services.prometheus.alertmanager;
+  mkConfigFile = pkgs.writeText "alertmanager.yml" (builtins.toJSON cfg.configuration);
+in {
+  options = {
+    services.prometheus.alertmanager = {
+      enable = mkEnableOption "Prometheus Alertmanager";
+
+      user = mkOption {
+        type = types.str;
+        default = "nobody";
+        description = ''
+          User name under which Alertmanager shall be run.
+        '';
+      };
+
+      group = mkOption {
+        type = types.str;
+        default = "nogroup";
+        description = ''
+          Group under which Alertmanager shall be run.
+        '';
+      };
+
+      configuration = mkOption {
+        type = types.attrs;
+        default = {};
+        description = ''
+          Alertmanager configuration as nix attribute set.
+        '';
+      };
+
+      logFormat = mkOption {
+        type = types.nullOr types.str;
+        default = null;
+        description = ''
+          If set use a syslog logger or JSON logging.
+        '';
+      };
+
+      logLevel = mkOption {
+        type = types.enum ["debug" "info" "warn" "error" "fatal"];
+        default = "warn";
+        description = ''
+          Only log messages with the given severity or above.
+        '';
+      };
+
+      webExternalUrl = mkOption {
+        type = types.nullOr types.str;
+        default = null;
+        description = ''
+          The URL under which Alertmanager is externally reachable (for example, if Alertmanager is served via a reverse proxy).
+          Used for generating relative and absolute links back to Alertmanager itself.
+          If the URL has a path portion, it will be used to prefix all HTTP endoints served by Alertmanager.
+          If omitted, relevant URL components will be derived automatically.
+        '';
+      };
+
+      listenAddress = mkOption {
+        type = types.nullOr types.str;
+        default = null;
+        description = ''
+          Address to listen on for the web interface and API.
+        '';
+      };
+
+      port = mkOption {
+        type = types.int;
+        default = 9093;
+        description = ''
+          Port to listen on for the web interface and API.
+        '';
+      };
+
+      openFirewall = mkOption {
+        type = types.bool;
+        default = false;
+        description = ''
+          Open port in firewall for incoming connections.
+        '';
+      };
+    };
+  };
+
+
+  config = mkIf cfg.enable {
+    networking.firewall.allowedTCPPorts = optional cfg.openFirewall cfg.port;
+
+    systemd.services.alertmanager = {
+      wantedBy = [ "multi-user.target" ];
+      after    = [ "network.target" ];
+      script = ''
+        ${pkgs.prometheus-alertmanager.bin}/bin/alertmanager \
+        -config.file ${mkConfigFile} \
+        -web.listen-address ${cfg.listenAddress}:${toString cfg.port} \
+        -log.level ${cfg.logLevel} \
+        ${optionalString (cfg.webExternalUrl != null) ''-web.external-url ${cfg.webExternalUrl} \''}
+        ${optionalString (cfg.logFormat != null) "-log.format ${cfg.logFormat}"}
+      '';
+
+      serviceConfig = {
+        User = cfg.user;
+        Group = cfg.group;
+        Restart  = "always";
+        PrivateTmp = true;
+        WorkingDirectory = "/tmp";
+        ExecReload = "${pkgs.coreutils}/bin/kill -HUP $MAINPID";
+      };
+    };
+  };
+}
diff --git a/nixos/modules/services/monitoring/prometheus/default.nix b/nixos/modules/services/monitoring/prometheus/default.nix
index 31979d2660c..e6817ee227a 100644
--- a/nixos/modules/services/monitoring/prometheus/default.nix
+++ b/nixos/modules/services/monitoring/prometheus/default.nix
@@ -29,6 +29,9 @@ let
     "-storage.local.path=${cfg.dataDir}/metrics"
     "-config.file=${writePrettyJSON "prometheus.yml" promConfig}"
     "-web.listen-address=${cfg.listenAddress}"
+    "-alertmanager.notification-queue-capacity=${toString cfg.alertmanagerNotificationQueueCapacity}"
+    "-alertmanager.timeout=${toString cfg.alertmanagerTimeout}s"
+    (optionalString (cfg.alertmanagerURL != []) "-alertmanager.url=${concatStringsSep "," cfg.alertmanagerURL}")
   ];
 
   promTypes.globalConfig = types.submodule {
@@ -388,6 +391,30 @@ in {
           A list of scrape configurations.
         '';
       };
+
+      alertmanagerURL = mkOption {
+        type = types.listOf types.str;
+        default = [];
+        description = ''
+          List of Alertmanager URLs to send notifications to.
+        '';
+      };
+
+      alertmanagerNotificationQueueCapacity = mkOption {
+        type = types.int;
+        default = 10000;
+        description = ''
+          The capacity of the queue for pending alert manager notifications.
+        '';
+      };
+
+      alertmanagerTimeout = mkOption {
+        type = types.int;
+        default = 10;
+        description = ''
+          Alert manager HTTP API timeout (in seconds).
+        '';
+      };
     };
   };