something

Signed-off-by: Cuong Nguyen <can@anyscale.com>

Author: can-anyscale

BUILD.bazel

@@ -119,6 +119,7 @@ ray_cc_library(
         ],
     ),
     deps = [
+        "//src/ray/observability:fake_metric",
         "//src/ray/observability:fake_ray_event_recorder",
     ],
 )

src/mock/ray/gcs/gcs_actor_manager.h

@@ -17,6 +17,7 @@
 #include <gmock/gmock.h>
 
 #include "ray/gcs/gcs_actor_manager.h"
+#include "ray/observability/fake_metric.h"
 #include "ray/observability/fake_ray_event_recorder.h"
 
 namespace ray {
@@ -38,7 +39,9 @@ class MockGcsActorManager : public GcsActorManager {
             [](const ActorID &) {},
             worker_client_pool,
             /*ray_event_recorder=*/fake_ray_event_recorder_,
-            /*session_name=*/"") {}
+            /*session_name=*/"",
+            /*actor_by_state_gauge=*/fake_actor_by_state_gauge_,
+            /*gcs_actor_by_state_gauge=*/fake_gcs_actor_by_state_gauge_) {}
 
   MOCK_METHOD(void,
               HandleRegisterActor,
@@ -85,6 +88,8 @@ class MockGcsActorManager : public GcsActorManager {
 
   instrumented_io_context mock_io_context_do_not_use_;
   observability::FakeRayEventRecorder fake_ray_event_recorder_;
+  observability::FakeGauge fake_actor_by_state_gauge_;
+  observability::FakeGauge fake_gcs_actor_by_state_gauge_;
 };
 
 }  // namespace gcs

src/mock/ray/gcs/gcs_placement_group_manager.h

@@ -17,14 +17,27 @@
 #include <gmock/gmock.h>
 
 #include "ray/gcs/gcs_placement_group_manager.h"
+#include "ray/observability/fake_metric.h"
 
 namespace ray {
 namespace gcs {
 
 class MockGcsPlacementGroupManager : public GcsPlacementGroupManager {
  public:
-  explicit MockGcsPlacementGroupManager(GcsResourceManager &gcs_resource_manager)
-      : GcsPlacementGroupManager(context_, gcs_resource_manager) {}
+  explicit MockGcsPlacementGroupManager(
+      GcsResourceManager &gcs_resource_manager,
+      ray::observability::MetricInterface &placement_group_gauge,
+      ray::observability::MetricInterface
+          &placement_group_creation_latency_in_ms_histogram,
+      ray::observability::MetricInterface
+          &placement_group_scheduling_latency_in_ms_histogram,
+      ray::observability::MetricInterface &placement_group_count_gauge)
+      : GcsPlacementGroupManager(context_,
+                                 gcs_resource_manager,
+                                 placement_group_gauge,
+                                 placement_group_creation_latency_in_ms_histogram,
+                                 placement_group_scheduling_latency_in_ms_histogram,
+                                 placement_group_count_gauge) {}
   MOCK_METHOD(void,
               HandleCreatePlacementGroup,
               (rpc::CreatePlacementGroupRequest request,

src/ray/common/BUILD.bazel

@@ -428,4 +428,4 @@ ray_cc_library(
     deps = [
         "//src/ray/stats:stats_lib",
     ],
-)
+)

src/ray/common/metrics.h

@@ -18,15 +18,16 @@
 
 namespace ray {
 
-inline ray::stats::Gauge GetActorMetric() {
+inline ray::stats::Gauge GetActorByStateGaugeMetric() {
   /// Tracks actors by state, including pending, running, and idle actors.
   ///
   /// To avoid metric collection conflicts between components reporting on the same actor,
   /// we use the "Source" required label.
   return ray::stats::Gauge{
       /*name=*/"actors",
-      /*description=*/"An actor can be in one of DEPENDENCIES_UNREADY, PENDING_CREATION, ALIVE, "
-      "ALIVE_IDLE, ALIVE_RUNNING_TASKS, RESTARTING, or DEAD states. ",
+      /*description=*/
+      "An actor can be in one of DEPENDENCIES_UNREADY, PENDING_CREATION, ALIVE, "
+      "ALIVE_IDLE, ALIVE_RUNNING_TASKS, RESTARTING, or DEAD states. "
       "An actor is considered ALIVE_IDLE if it is not executing any tasks.",
       /*unit=*/"",
       // State: the actor state, which is from rpc::ActorTableData::ActorState,

src/ray/core_worker/BUILD.bazel

@@ -32,8 +32,8 @@ ray_cc_library(
         ":reference_count",
         ":shutdown_coordinator",
         ":task_event_buffer",
-        "//src/ray/common:protobuf_utils",
         "//src/ray/common:metrics",
+        "//src/ray/common:protobuf_utils",
         "//src/ray/core_worker/task_execution:task_receiver",
         "//src/ray/core_worker/task_submission:normal_task_submitter",
         "//src/ray/core_worker_rpc_client:core_worker_client",

src/ray/core_worker/core_worker.cc

@@ -228,16 +228,16 @@ void TaskCounter::RecordMetrics() {
     } else {
       running_tasks = 1.0;
     }
-    ray::stats::STATS_actors.Record(idle,
-                                    {{"State", "ALIVE_IDLE"},
-                                     {"Name", actor_name_},
-                                     {"Source", "executor"},
-                                     {"JobId", job_id_}});
-    ray::stats::STATS_actors.Record(running_tasks,
-                                    {{"State", "ALIVE_RUNNING_TASKS"},
-                                     {"Name", actor_name_},
-                                     {"Source", "executor"},
-                                     {"JobId", job_id_}});
+    actor_by_state_gauge_.Record(idle,
+                                 {{"State"sv, "ALIVE_IDLE"},
+                                  {"Name"sv, actor_name_},
+                                  {"Source"sv, "executor"},
+                                  {"JobId"sv, job_id_}});
+    actor_by_state_gauge_.Record(running_tasks,
+                                 {{"State"sv, "ALIVE_RUNNING_TASKS"},
+                                  {"Name"sv, actor_name_},
+                                  {"Source"sv, "executor"},
+                                  {"JobId"sv, job_id_}});
   }
 }
 

src/ray/core_worker/core_worker_process.h

@@ -185,8 +185,8 @@ class CoreWorkerProcessImpl {
   /// The client to export metrics to the metrics agent.
   std::unique_ptr<ray::rpc::MetricsAgentClient> metrics_agent_client_;
 
-  ray::stats::Gauge task_by_state_gauge_{GetTaskMetric()};
-  ray::stats::Gauge actor_by_state_gauge_{GetActorMetric()};
+  ray::stats::Gauge task_by_state_gauge_{GetTaskByStateGaugeMetric()};
+  ray::stats::Gauge actor_by_state_gauge_{GetActorByStateGaugeMetric()};
 };
 }  // namespace core
 }  // namespace ray

src/ray/core_worker/metrics.h

@@ -19,7 +19,7 @@
 namespace ray {
 namespace core {
 
-inline ray::stats::Gauge GetTaskMetric() {
+inline ray::stats::Gauge GetTaskByStateGaugeMetric() {
   /// Tracks tasks by state, including pending, running, and finished tasks.
   /// This metric may be recorded from multiple components processing the task in Ray,
   /// including the submitting core worker, executor core worker, and pull manager.

src/ray/core_worker/tests/core_worker_test.cc

@@ -291,8 +291,8 @@ class CoreWorkerTest : public ::testing::Test {
   pubsub::Publisher *object_info_publisher_;
   std::shared_ptr<TaskManager> task_manager_;
   std::shared_ptr<CoreWorker> core_worker_;
-  ray::observability::FakeMetric fake_task_by_state_gauge_;
-  ray::observability::FakeMetric fake_actor_by_state_gauge_;
+  ray::observability::FakeGauge fake_task_by_state_gauge_;
+  ray::observability::FakeGauge fake_actor_by_state_gauge_;
   std::unique_ptr<FakePeriodicalRunner> fake_periodical_runner_;
 
   // Controllable time for testing publisher timeouts