Merge pull request #1173 from DeterminateSystems/queue-runner-exporter

hydra-queue-runner metrics

doc/manual/src/configuration.md

@@ -102,6 +102,22 @@ in the hydra configuration file, as below:
 </hydra_notify>
 ```
 
+hydra-queue-runner's Prometheus service
+---------------------------------------
+
+hydra-queue-runner supports running a Prometheus webserver for metrics. The
+exporter's address defaults to exposing on `127.0.0.1:9198`, but is also
+configurable through the hydra configuration file and a command line argument,
+as below. A port of `:0` will make the exposer choose a random, available port.
+
+```conf
+queue_runner_exporter_address = 127.0.0.1:9198
+```
+
+```shell
+$ hydra-queue-runner --prometheus-address 127.0.0.1:9198
+```
+
 Using LDAP as authentication backend (optional)
 -----------------------------------------------
 

flake.lock

@@ -16,6 +16,22 @@
         "type": "github"
       }
     },
+    "newNixpkgs": {
+      "locked": {
+        "lastModified": 1647380550,
+        "narHash": "sha256-909TI9poX7CIUiFx203WL29YON6m/I6k0ExbZvR7bLM=",
+        "owner": "NixOS",
+        "repo": "nixpkgs",
+        "rev": "6e3ee8957637a60f5072e33d78e05c0f65c54366",
+        "type": "github"
+      },
+      "original": {
+        "owner": "NixOS",
+        "ref": "nixos-unstable-small",
+        "repo": "nixpkgs",
+        "type": "github"
+      }
+    },
     "nix": {
       "inputs": {
         "lowdown-src": "lowdown-src",
@@ -67,6 +83,7 @@
     },
     "root": {
       "inputs": {
+        "newNixpkgs": "newNixpkgs",
         "nix": "nix",
         "nixpkgs": [
           "nix",

flake.nix

@@ -1,10 +1,13 @@
 {
   description = "A Nix-based continuous build system";
 
+  # FIXME: All the pinned versions of nix/nixpkgs have a broken foreman (yes,
+  # even 2.7.0's Nixpkgs pin).
+  inputs.newNixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable-small";
   inputs.nixpkgs.follows = "nix/nixpkgs";
   #inputs.nix.url = github:NixOS/nix/2.7.0;
 
-  outputs = { self, nixpkgs, nix }:
+  outputs = { self, newNixpkgs, nixpkgs, nix }:
     let
 
       version = "${builtins.readFile ./version.txt}.${builtins.substring 0 8 (self.lastModifiedDate or "19700101")}.${self.shortRev or "DIRTY"}";
@@ -38,6 +41,12 @@
       # A Nixpkgs overlay that provides a 'hydra' package.
       overlay = final: prev: {
 
+        # Overlay these packages to use dependencies from the Nixpkgs everything
+        # else uses, to side-step the version difference: glibc is 2.32 in the
+        # nix-pinned Nixpkgs, but 2.33 in the newNixpkgs commit.
+        civetweb = final.callPackage "${newNixpkgs}/pkgs/development/libraries/civetweb" { };
+        prometheus-cpp = final.callPackage "${newNixpkgs}/pkgs/development/libraries/prometheus-cpp" { };
+
         # Add LDAP dependencies that aren't currently found within nixpkgs.
         perlPackages = prev.perlPackages // {
           TestPostgreSQL = final.perlPackages.buildPerlModule {
@@ -61,7 +70,7 @@
             };
           };
 
-          FunctionParameters = final.buildPerlPackage {
+          FunctionParameters = final.perlPackages.buildPerlPackage {
             pname = "Function-Parameters";
             version = "2.001003";
             src = final.fetchurl {
@@ -75,7 +84,7 @@
             };
           };
 
-          CatalystPluginPrometheusTiny = final.buildPerlPackage {
+          CatalystPluginPrometheusTiny = final.perlPackages.buildPerlPackage {
             pname = "Catalyst-Plugin-PrometheusTiny";
             version = "0.005";
             src = final.fetchurl {
@@ -104,7 +113,7 @@
             };
           };
 
-          CryptPassphrase = final.buildPerlPackage {
+          CryptPassphrase = final.perlPackages.buildPerlPackage {
             pname = "Crypt-Passphrase";
             version = "0.003";
             src = final.fetchurl {
@@ -117,7 +126,7 @@
             };
           };
 
-          CryptPassphraseArgon2 = final.buildPerlPackage {
+          CryptPassphraseArgon2 = final.perlPackages.buildPerlPackage {
             pname = "Crypt-Passphrase-Argon2";
             version = "0.002";
             src = final.fetchurl {
@@ -131,7 +140,7 @@
             };
           };
 
-          DataRandom = final.buildPerlPackage {
+          DataRandom = final.perlPackages.buildPerlPackage {
             pname = "Data-Random";
             version = "0.13";
             src = final.fetchurl {
@@ -145,7 +154,7 @@
             };
           };
 
-          DirSelf = final.buildPerlPackage {
+          DirSelf = final.perlPackages.buildPerlPackage {
             pname = "Dir-Self";
             version = "0.11";
             src = final.fetchurl {
@@ -173,7 +182,7 @@
             };
           };
 
-          PrometheusTiny = final.buildPerlPackage {
+          PrometheusTiny = final.perlPackages.buildPerlPackage {
             pname = "Prometheus-Tiny";
             version = "0.007";
             src = final.fetchurl {
@@ -188,7 +197,7 @@
             };
           };
 
-          PrometheusTinyShared = final.buildPerlPackage {
+          PrometheusTinyShared = final.perlPackages.buildPerlPackage {
             pname = "Prometheus-Tiny-Shared";
             version = "0.023";
             src = final.fetchurl {
@@ -219,7 +228,7 @@
             };
           };
 
-          TieHashMethod = final.buildPerlPackage {
+          TieHashMethod = final.perlPackages.buildPerlPackage {
             pname = "Tie-Hash-Method";
             version = "0.02";
             src = final.fetchurl {
@@ -232,7 +241,7 @@
             };
           };
 
-          Test2Harness = final.buildPerlPackage {
+          Test2Harness = final.perlPackages.buildPerlPackage {
             pname = "Test2-Harness";
             version = "1.000042";
             src = final.fetchurl {
@@ -281,7 +290,7 @@
             };
           };
 
-          LongJump = final.buildPerlPackage {
+          LongJump = final.perlPackages.buildPerlPackage {
             pname = "Long-Jump";
             version = "0.000001";
             src = final.fetchurl {
@@ -295,7 +304,7 @@
             };
           };
 
-          gotofile = final.buildPerlPackage {
+          gotofile = final.perlPackages.buildPerlPackage {
             pname = "goto-file";
             version = "0.005";
             src = final.fetchurl {
@@ -425,7 +434,7 @@
             };
           };
 
-          StringCompareConstantTime = final.buildPerlPackage {
+          StringCompareConstantTime = final.perlPackages.buildPerlPackage {
             pname = "String-Compare-ConstantTime";
             version = "0.321";
             src = final.fetchurl {
@@ -438,7 +447,7 @@
             };
           };
 
-          UUID4Tiny = final.buildPerlPackage {
+          UUID4Tiny = final.perlPackages.buildPerlPackage {
             pname = "UUID4-Tiny";
             version = "0.002";
             src = final.fetchurl {
@@ -562,13 +571,16 @@
               (if lib.versionAtLeast lib.version "20.03pre"
               then nlohmann_json
               else nlohmann_json.override { multipleHeaders = true; })
+              prometheus-cpp
             ];
 
           checkInputs = [
             cacert
-            foreman
+            # FIXME: foreman is broken on all nix/nixpkgs pin, up to and
+            # including 2.7.0
+            newNixpkgs.legacyPackages.${final.system}.foreman
             glibcLocales
-            netcat-openbsd
+            libressl.nc
             openldap
             python3
           ];

src/hydra-queue-runner/Makefile.am

@@ -4,5 +4,5 @@ hydra_queue_runner_SOURCES = hydra-queue-runner.cc queue-monitor.cc dispatcher.c
  builder.cc build-result.cc build-remote.cc \
  hydra-build-result.hh counter.hh state.hh db.hh \
  nar-extractor.cc nar-extractor.hh
-hydra_queue_runner_LDADD = $(NIX_LIBS) -lpqxx
+hydra_queue_runner_LDADD = $(NIX_LIBS) -lpqxx -lprometheus-cpp-pull -lprometheus-cpp-core
 hydra_queue_runner_CXXFLAGS = $(NIX_CFLAGS) -Wall -I ../libhydra -Wno-deprecated-declarations

src/hydra-queue-runner/hydra-queue-runner.cc

@@ -6,6 +6,8 @@
 #include <sys/stat.h>
 #include <fcntl.h>
 
+#include <prometheus/exposer.h>
+
 #include "state.hh"
 #include "hydra-build-result.hh"
 #include "store-api.hh"
@@ -36,8 +38,55 @@ std::string getEnvOrDie(const std::string & key)
     return *value;
 }
 
+State::PromMetrics::PromMetrics()
+    : registry(std::make_shared<prometheus::Registry>())
+    , queue_checks_started(
+        prometheus::BuildCounter()
+            .Name("hydraqueuerunner_queue_checks_started_total")
+            .Help("Number of times State::getQueuedBuilds() was started")
+            .Register(*registry)
+            .Add({})
+    )
+    , queue_build_loads(
+        prometheus::BuildCounter()
+            .Name("hydraqueuerunner_queue_build_loads_total")
+            .Help("Number of builds loaded")
+            .Register(*registry)
+            .Add({})
+    )
+    , queue_steps_created(
+        prometheus::BuildCounter()
+            .Name("hydraqueuerunner_queue_steps_created_total")
+            .Help("Number of steps created")
+            .Register(*registry)
+            .Add({})
+    )
+    , queue_checks_early_exits(
+        prometheus::BuildCounter()
+            .Name("hydraqueuerunner_queue_checks_early_exits_total")
+            .Help("Number of times State::getQueuedBuilds() yielded to potential bumps")
+            .Register(*registry)
+            .Add({})
+    )
+    , queue_checks_finished(
+        prometheus::BuildCounter()
+            .Name("hydraqueuerunner_queue_checks_finished_total")
+            .Help("Number of times State::getQueuedBuilds() was completed")
+            .Register(*registry)
+            .Add({})
+    )
+    , queue_max_id(
+        prometheus::BuildGauge()
+            .Name("hydraqueuerunner_queue_max_build_id_info")
+            .Help("Maximum build record ID in the queue")
+            .Register(*registry)
+            .Add({})
+    )
+{
 
-State::State()
+}
+
+State::State(std::optional<std::string> metricsAddrOpt)
     : config(std::make_unique<HydraConfig>())
     , maxUnsupportedTime(config->getIntOption("max_unsupported_time", 0))
     , dbPool(config->getIntOption("max_db_connections", 128))
@@ -45,11 +94,16 @@ State::State()
     , maxLogSize(config->getIntOption("max_log_size", 64ULL << 20))
     , uploadLogsToBinaryCache(config->getBoolOption("upload_logs_to_binary_cache", false))
     , rootsDir(config->getStrOption("gc_roots_dir", fmt("%s/gcroots/per-user/%s/hydra-roots", settings.nixStateDir, getEnvOrDie("LOGNAME"))))
+    , metricsAddr(config->getStrOption("queue_runner_metrics_address", std::string{"127.0.0.1:9198"}))
 {
     hydraData = getEnvOrDie("HYDRA_DATA");
 
     logDir = canonPath(hydraData + "/build-logs");
 
+    if (metricsAddrOpt.has_value()) {
+        metricsAddr = metricsAddrOpt.value();
+    }
+
     /* handle deprecated store specification */
     if (config->getStrOption("store_mode") != "")
         throw Error("store_mode in hydra.conf is deprecated, please use store_uri");
@@ -754,6 +808,18 @@ void State::run(BuildID buildOne)
     if (!lock)
         throw Error("hydra-queue-runner is already running");
 
+    std::cout << "Starting the Prometheus exporter on " << metricsAddr << std::endl;
+
+    /* Set up simple exporter, to show that we're still alive. */
+    prometheus::Exposer promExposer{metricsAddr};
+    auto exposerPort = promExposer.GetListeningPorts().front();
+
+    promExposer.RegisterCollectable(prom.registry);
+
+    std::cout << "Started the Prometheus exporter, listening on "
+        << metricsAddr << "/metrics (port " << exposerPort << ")"
+        << std::endl;
+
     Store::Params localParams;
     localParams["max-connections"] = "16";
     localParams["max-connection-age"] = "600";
@@ -864,6 +930,7 @@ int main(int argc, char * * argv)
         bool unlock = false;
         bool status = false;
         BuildID buildOne = 0;
+        std::optional<std::string> metricsAddrOpt = std::nullopt;
 
         parseCmdLine(argc, argv, [&](Strings::iterator & arg, const Strings::iterator & end) {
             if (*arg == "--unlock")
@@ -875,6 +942,8 @@ int main(int argc, char * * argv)
                     buildOne = *b;
                 else
                     throw Error("‘--build-one’ requires a build ID");
+            } else if (*arg == "--prometheus-address") {
+                metricsAddrOpt = getArg(*arg, arg, end);
             } else
                 return false;
             return true;
@@ -883,7 +952,7 @@ int main(int argc, char * * argv)
         settings.verboseBuild = true;
         settings.lockCPU = false;
 
-        State state;
+        State state{metricsAddrOpt};
         if (status)
             state.showStatus();
         else if (unlock)

src/hydra-queue-runner/queue-monitor.cc

@@ -82,6 +82,8 @@ struct PreviousFailure : public std::exception {
 bool State::getQueuedBuilds(Connection & conn,
     ref<Store> destStore, unsigned int & lastBuildId)
 {
+    prom.queue_checks_started.Increment();
+
     printInfo("checking the queue for builds > %d...", lastBuildId);
 
     /* Grab the queued builds from the database, but don't process
@@ -107,7 +109,10 @@ bool State::getQueuedBuilds(Connection & conn,
             auto builds_(builds.lock());
             BuildID id = row["id"].as<BuildID>();
             if (buildOne && id != buildOne) continue;
-            if (id > newLastBuildId) newLastBuildId = id;
+            if (id > newLastBuildId) {
+                newLastBuildId = id;
+                prom.queue_max_id.Set(id);
+            }
             if (builds_->count(id)) continue;
 
             auto build = std::make_shared<Build>(
@@ -136,6 +141,7 @@ bool State::getQueuedBuilds(Connection & conn,
     std::set<StorePath> finishedDrvs;
 
     createBuild = [&](Build::ptr build) {
+        prom.queue_build_loads.Increment();
         printMsg(lvlTalkative, format("loading build %1% (%2%)") % build->id % build->fullJobName());
         nrAdded++;
         newBuildsByID.erase(build->id);
@@ -306,8 +312,13 @@ bool State::getQueuedBuilds(Connection & conn,
 
         /* Stop after a certain time to allow priority bumps to be
            processed. */
-        if (std::chrono::system_clock::now() > start + std::chrono::seconds(600)) break;
+        if (std::chrono::system_clock::now() > start + std::chrono::seconds(600)) {
+            prom.queue_checks_early_exits.Increment();
+            break;
         } 
+    }
+
+    prom.queue_checks_finished.Increment();
 
     lastBuildId = newBuildsByID.empty() ? newLastBuildId : newBuildsByID.begin()->first - 1;
     return newBuildsByID.empty();
@@ -437,6 +448,8 @@ Step::ptr State::createStep(ref<Store> destStore,
 
     if (!isNew) return step;
 
+    prom.queue_steps_created.Increment();
+
     printMsg(lvlDebug, "considering derivation ‘%1%’", localStore->printStorePath(drvPath));
 
     /* Initialize the step. Note that the step may be visible in

src/hydra-queue-runner/state.hh

@@ -7,6 +7,10 @@
 #include <memory>
 #include <queue>
 
+#include <prometheus/counter.h>
+#include <prometheus/gauge.h>
+#include <prometheus/registry.h>
+
 #include "db.hh"
 
 #include "parsed-derivations.hh"
@@ -433,8 +437,25 @@ private:
        via gc_roots_dir. */
     nix::Path rootsDir;
 
+    std::string metricsAddr;
+
+    struct PromMetrics
+    {
+        std::shared_ptr<prometheus::Registry> registry;
+
+        prometheus::Counter& queue_checks_started;
+        prometheus::Counter& queue_build_loads;
+        prometheus::Counter& queue_steps_created;
+        prometheus::Counter& queue_checks_early_exits;
+        prometheus::Counter& queue_checks_finished;
+        prometheus::Gauge& queue_max_id;
+
+        PromMetrics();
+    };
+    PromMetrics prom;
+
 public:
-    State();
+    State(std::optional<std::string> metricsAddrOpt);
 
 private:
 
@@ -544,6 +565,8 @@ private:
 
     void addRoot(const nix::StorePath & storePath);
 
+    void runMetricsExporter();
+
 public:
 
     void showStatus();

t/Hydra/Config/include.t

@@ -20,6 +20,7 @@ write_file($ctx{'tmpdir'} . "/bar.conf", q|
 |);
 
 is(getHydraConfig(), {
+    queue_runner_metrics_address => "127.0.0.1:0",
     foo => { bar => "baz" }
 }, "Nested includes work.");
 

t/lib/HydraTestContext.pm

@@ -51,6 +51,7 @@ sub new {
     $ENV{'HYDRA_CONFIG'} = "$dir/hydra.conf";
 
     my $hydra_config = $opts{'hydra_config'} || "";
+    $hydra_config = "queue_runner_metrics_address = 127.0.0.1:0\n" . $hydra_config;
     if ($opts{'use_external_destination_store'} // 1) {
         $hydra_config = "store_uri = file:$dir/nix/dest-store\n" . $hydra_config;
     }