Scheduler issues read the utilization fields the event carries, flagged the way sp_HealthParser flags them (#4452)

Darling/Darling.Tests/DarlingMcpHealthParserToolsTests.cs

@@ -208,12 +208,12 @@ viewer entry point still delegates rather than growing an independent (drift-pro
     [Fact]
     public void Significance_SchedulerIssue_MatchesViewer()
     {
-        foreach (var status in new[] { "WARNING", "OK", "", null })
+        foreach (var (sqlCpu, other, mem) in new (int?, int?, int?)[] { (90, 0, 100), (89, 0, 100), (0, 50, 100), (0, 49, 100), (0, 0, 50), (0, 0, 51), (null, null, null) })
         {
-            var r = new SchedulerIssueRecord { Status = status };
+            var r = new SchedulerIssueRecord { SqlCpuUtilization = sqlCpu, OtherProcessCpu = other, MemoryUtilization = mem };
             Assert.Equal(SystemEventSignificance.IsSignificant(r), SystemHealthSignificance.IsSignificant(r));
         }
-        Assert.True(SystemHealthSignificance.IsSignificant(new SchedulerIssueRecord { Status = "WARNING" }));
+        Assert.True(SystemHealthSignificance.IsSignificant(new SchedulerIssueRecord { SqlCpuUtilization = 90 }));
     }
 
     [Fact]
@@ -311,7 +311,7 @@ public void RoundTrip_ShredFixturesThenGate_KeepsOnlySignificantRows()
         // SystemHealthSignificance keeps the significant rows. Assert the same significant set the
         // viewer's System Events tab surfaces.
         Assert.True(SystemHealthSignificance.IsSignificant(
-            SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor.xml"))!));            // WARNING
+            SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor_high_sql_cpu.xml"))!));  // SQL CPU 94 >= 90
         Assert.True(SystemHealthSignificance.IsSignificant(
             SystemHealthParser.ParseSevereError(LoadFixture("error_reported.xml"))!));                  // severity 24
         Assert.True(SystemHealthSignificance.IsSignificant(
@@ -398,7 +398,7 @@ await DarlingMcpTestData.ExecAsync(connection, ct,
 VALUES ($1,$2,$3,$4,$5,$6,$7)",
                     CollectionIdGenerator.Next(), t, ServerId, ServerName, t, eventType, LoadFixture(fixture));
 
-            await PlantEvent(SystemHealthParser.SchedulerMonitorEvent, "scheduler_monitor.xml");
+            await PlantEvent(SystemHealthParser.SchedulerMonitorEvent, "scheduler_monitor_high_sql_cpu.xml");
             await PlantEvent(SystemHealthParser.ErrorReportedEvent, "error_reported.xml");
             await PlantEvent(SystemHealthParser.SpServerDiagnosticsEvent, "sp_server_diagnostics_system.xml");
             await PlantEvent(SystemHealthParser.SpServerDiagnosticsEvent, "sp_server_diagnostics_query_processing_warning.xml");
@@ -441,10 +441,20 @@ await DarlingMcpTestData.ExecAsync(connection, ct,
             Assert.Contains("the absence is a measurement", broker.GetProperty("message").GetString()!, StringComparison.Ordinal);
 
             /* The data envelope carries the same witness pair. */
-            var scheduler = JsonDocument.Parse(await DarlingMcpHealthParserTools.GetSchedulerIssues(postgres, ServerName)).RootElement;
+            var schedulerJson = await DarlingMcpHealthParserTools.GetSchedulerIssues(postgres, ServerName);
+            var scheduler = JsonDocument.Parse(schedulerJson).RootElement;
             Assert.True(scheduler.GetProperty("source_observed").GetBoolean());
             Assert.Equal(t.ToString("o"), scheduler.GetProperty("last_captured_at").GetString());
 
+            /* #4452: the utilization fields the planted high-CPU sample carries — SQL CPU pinned at 94,
+               other-process CPU 4, idle 2, memory 100 — come back on the tool's own output text. This is
+               the port's product-path pin: it asserts on the served JSON text (not typed properties) so it
+               also compiles, unchanged, against the pre-port code, where these fields are null or absent. */
+            Assert.Contains("\"sql_cpu_utilization\":94", schedulerJson, StringComparison.Ordinal);
+            Assert.Contains("\"other_process_cpu\":4", schedulerJson, StringComparison.Ordinal);
+            Assert.Contains("\"system_idle\":2", schedulerJson, StringComparison.Ordinal);
+            Assert.Contains("\"memory_utilization\":100", schedulerJson, StringComparison.Ordinal);
+
             /* an unknown server resolves to the listing error. */
             Assert.StartsWith("Could not resolve server.", McpHelpers.ErrorMessageOf(await DarlingMcpHealthParserTools.GetSystemHealth(postgres, "darling-no-such-server")), StringComparison.Ordinal);
 

Darling/Darling.Tests/Fixtures/SystemHealth/scheduler_monitor_high_sql_cpu.xml

@@ -0,0 +1,16 @@
+<!-- #4452: real scheduler_monitor_system_health_ring_buffer_recorded event from a monitored field
+     store, call_stack stripped. process_utilization=94, system_idle=73->2 (other-process CPU =
+     100-94-2=4), memory_utilization=100, page_faults=505, working_set_delta=4096 bytes ->
+     4096/1048576 = 0.0039... MB, rounded to 0.00. SqlCpuUtilization 94 >= 90, so this sample is
+     significant on the SQL-CPU arm. -->
+<event name="scheduler_monitor_system_health_ring_buffer_recorded" package="sqlos" timestamp="2026-09-26T22:08:48.939Z">
+  <data name="id"><type name="uint32" package="package0"/><value>0</value></data>
+  <data name="timestamp"><type name="uint64" package="package0"/><value>0</value></data>
+  <data name="process_utilization"><type name="uint32" package="package0"/><value>94</value></data>
+  <data name="system_idle"><type name="uint32" package="package0"/><value>2</value></data>
+  <data name="user_mode_time"><type name="uint64" package="package0"/><value>2233593750</value></data>
+  <data name="kernel_mode_time"><type name="uint64" package="package0"/><value>30312500</value></data>
+  <data name="page_faults"><type name="uint32" package="package0"/><value>505</value></data>
+  <data name="working_set_delta"><type name="int64" package="package0"/><value>4096</value></data>
+  <data name="memory_utilization"><type name="uint32" package="package0"/><value>100</value></data>
+</event>

Darling/Darling.Tests/Fixtures/SystemHealth/scheduler_monitor_low_memory.xml

@@ -0,0 +1,16 @@
+<!-- #4452: DERIVED from the real scheduler_monitor_normal.xml capture (same element shape; only
+     the memory_utilization <value> number was changed). No real captured sample crossed the
+     low-memory threshold, so this fixture exercises that arm: sql_cpu=3, system_idle=95 ->
+     other-process CPU = 100-3-95=2, memory_utilization=50, which is <= 50 and so significant on
+     the memory arm alone (sql_cpu 3 < 90, other-process CPU 2 < 50). -->
+<event name="scheduler_monitor_system_health_ring_buffer_recorded" package="sqlos" timestamp="2026-09-26T23:59:53.232Z">
+  <data name="id"><type name="uint32" package="package0"/><value>0</value></data>
+  <data name="timestamp"><type name="uint64" package="package0"/><value>0</value></data>
+  <data name="process_utilization"><type name="uint32" package="package0"/><value>3</value></data>
+  <data name="system_idle"><type name="uint32" package="package0"/><value>95</value></data>
+  <data name="user_mode_time"><type name="uint64" package="package0"/><value>176250000</value></data>
+  <data name="kernel_mode_time"><type name="uint64" package="package0"/><value>7187500</value></data>
+  <data name="page_faults"><type name="uint32" package="package0"/><value>99</value></data>
+  <data name="working_set_delta"><type name="int64" package="package0"/><value>77824</value></data>
+  <data name="memory_utilization"><type name="uint32" package="package0"/><value>50</value></data>
+</event>

Darling/Darling.Tests/Fixtures/SystemHealth/scheduler_monitor_normal.xml

@@ -0,0 +1,16 @@
+<!-- #4452: real scheduler_monitor_system_health_ring_buffer_recorded event from a monitored field
+     store, call_stack stripped. process_utilization=3, system_idle=95 -> other-process CPU =
+     100-3-95=2, memory_utilization=100, page_faults=99, working_set_delta=77824 bytes ->
+     77824/1048576 = 0.0742... MB, rounded to 0.07. Every arm is well inside its threshold
+     (3 < 90, 2 < 50, 100 > 50), so this sample is NOT significant. -->
+<event name="scheduler_monitor_system_health_ring_buffer_recorded" package="sqlos" timestamp="2026-09-26T23:59:53.232Z">
+  <data name="id"><type name="uint32" package="package0"/><value>0</value></data>
+  <data name="timestamp"><type name="uint64" package="package0"/><value>0</value></data>
+  <data name="process_utilization"><type name="uint32" package="package0"/><value>3</value></data>
+  <data name="system_idle"><type name="uint32" package="package0"/><value>95</value></data>
+  <data name="user_mode_time"><type name="uint64" package="package0"/><value>176250000</value></data>
+  <data name="kernel_mode_time"><type name="uint64" package="package0"/><value>7187500</value></data>
+  <data name="page_faults"><type name="uint32" package="package0"/><value>99</value></data>
+  <data name="working_set_delta"><type name="int64" package="package0"/><value>77824</value></data>
+  <data name="memory_utilization"><type name="uint32" package="package0"/><value>100</value></data>
+</event>

Darling/Darling.Tests/Fixtures/SystemHealth/scheduler_monitor_other_process_cpu.xml

@@ -0,0 +1,16 @@
+<!-- #4452: DERIVED from the real scheduler_monitor_normal.xml capture (same element shape;
+     only the process_utilization and system_idle <value> numbers were changed). No real captured
+     sample crossed the other-process-CPU threshold, so this fixture exercises that arm: sql_cpu=3,
+     system_idle=40 -> other-process CPU = 100-3-40=57, which is >= 50 and so significant on the
+     other-process arm alone (sql_cpu 3 < 90, memory_utilization 100 > 50). -->
+<event name="scheduler_monitor_system_health_ring_buffer_recorded" package="sqlos" timestamp="2026-09-26T23:59:53.232Z">
+  <data name="id"><type name="uint32" package="package0"/><value>0</value></data>
+  <data name="timestamp"><type name="uint64" package="package0"/><value>0</value></data>
+  <data name="process_utilization"><type name="uint32" package="package0"/><value>3</value></data>
+  <data name="system_idle"><type name="uint32" package="package0"/><value>40</value></data>
+  <data name="user_mode_time"><type name="uint64" package="package0"/><value>176250000</value></data>
+  <data name="kernel_mode_time"><type name="uint64" package="package0"/><value>7187500</value></data>
+  <data name="page_faults"><type name="uint32" package="package0"/><value>99</value></data>
+  <data name="working_set_delta"><type name="int64" package="package0"/><value>77824</value></data>
+  <data name="memory_utilization"><type name="uint32" package="package0"/><value>100</value></data>
+</event>

Darling/Darling.Tests/McpToolGuideHeads.HealthParser.cs

@@ -42,7 +42,7 @@ private static readonly (string Tool, string Fact)[] GateFacts =
         ("get_health_parser_memory_node_oom", "Ungated: every recorded OOM is returned."),
         ("get_health_parser_severe_errors", "Gated: severity 19 or higher only, benign connection-reset error numbers excluded"),
         ("get_health_parser_io_issues", "Gated: WARNING-state results only."),
-        ("get_health_parser_scheduler_issues", "Gated: WARNING-state results only."),
+        ("get_health_parser_scheduler_issues", "Gated: only samples where SQL Server's own CPU is pinned, other processes are using half the box or more, or memory utilization has dropped to half or below."),
         ("get_health_parser_memory_conditions", "Gated: only snapshots whose last notification is RESOURCE_MEMPHYSICAL_LOW."),
         ("get_health_parser_cpu_tasks", "Gated: WARNING-state results with at least 10 pending tasks only."),
         ("get_health_parser_memory_broker", "Gated: RESOURCE_MEMPHYSICAL_LOW notifications only."),

Darling/Darling.Tests/McpToolsListBudget/DarlingMcpHealthParserTools.txt

@@ -29,7 +29,7 @@ param get_health_parser_memory_node_oom.hours_back 41
 param get_health_parser_memory_node_oom.limit 38
 param get_health_parser_memory_node_oom.server_name 28
 
-tool get_health_parser_scheduler_issues 446
+tool get_health_parser_scheduler_issues 536
 param get_health_parser_scheduler_issues.as_of 167
 param get_health_parser_scheduler_issues.hours_back 41
 param get_health_parser_scheduler_issues.limit 38

Darling/Darling.Tests/McpToolsListBudgetTests.cs

@@ -184,7 +184,10 @@ shared 32 KB budget and is not a served description either. */
        this PR's own get_store_host (+616) and dev's #4279 Lite-parity head change (+181). Constant set to the
        value McpToolsListBudgetTests itself measured on the merged tree, not the two deltas added by hand. */
     // #4442: get_read_latency's served block.
-    private const int TotalCeilingBytes = 176_182;
+    // #4452 (merge): re-measured on the tree combining dev's #4442 get_read_latency addition with this
+    // branch's scheduler-issues tool description growth (+95). Constant set to the value
+    // McpToolsListBudgetTests itself measured on the merged tree, not the two deltas added by hand.
+    private const int TotalCeilingBytes = 176_277;
 
 
 

Darling/Darling.Tests/SystemHealthParserTests.cs

@@ -27,7 +27,7 @@ public class SystemHealthParserTests
     private static string LoadFixture(string name) =>
         File.ReadAllText(Path.Combine(AppContext.BaseDirectory, "Fixtures", "SystemHealth", name));
 
-    private static readonly DateTime SchedulerTime = new(2026, 7, 5, 11, 58, 0, 123, DateTimeKind.Utc);
+    private static readonly DateTime SchedulerHighCpuTime = new(2026, 9, 26, 22, 8, 48, 939, DateTimeKind.Utc);
     private static readonly DateTime ErrorTime = new(2026, 7, 5, 12, 0, 5, 500, DateTimeKind.Utc);
     private static readonly DateTime ResourceTime = new(2026, 7, 5, 12, 1, 0, 0, DateTimeKind.Utc);
     private static readonly DateTime BrokerTime = new(2026, 7, 5, 12, 2, 10, 250, DateTimeKind.Utc);
@@ -41,20 +41,75 @@ private static string LoadFixture(string name) =>
     // ── SchedulerIssues (scheduler_monitor_system_health_ring_buffer_recorded) ──
 
     [Fact]
-    public void SchedulerIssue_ShredsEveryColumn()
+    public void SchedulerIssue_ShredsEveryColumn_HighSqlCpu()
     {
-        var r = SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor.xml"));
+        // Real field capture: process_utilization=94, system_idle=2 -> other = 100-94-2=4;
+        // working_set_delta 4096 bytes -> 4096/1048576 = 0.00390625 MB, rounded to 0.00.
+        var r = SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor_high_sql_cpu.xml"));
 
         Assert.NotNull(r);
-        Assert.Equal(SchedulerTime, r!.EventTime);
-        Assert.Equal(10, r.SchedulerId);
-        Assert.Equal(3, r.CpuId);
-        Assert.Equal("WARNING", r.Status);          // data[@name="status"]/text, NOT the numeric value
-        Assert.True(r.IsOnline);
-        Assert.False(r.IsRunnable);
-        Assert.True(r.IsRunning);
-        Assert.Equal(4500L, r.NonYieldingTimeMs);    // data[@name="non_yielding_time"] (no _ms suffix in XML)
-        Assert.Equal(4L, r.ThreadQuantumMs);         // data[@name="thread_quantum"]
+        Assert.Equal(SchedulerHighCpuTime, r!.EventTime);
+        Assert.Equal(94, r.SqlCpuUtilization);
+        Assert.Equal(4, r.OtherProcessCpu);
+        Assert.Equal(2, r.SystemIdle);
+        Assert.Equal(100, r.MemoryUtilization);
+        Assert.Equal(505L, r.PageFaults);
+        Assert.Equal(0.00m, r.WorkingSetDeltaMb);
+        Assert.True(SystemHealthSignificance.IsSignificant(r));   // SQL CPU 94 >= 90
+    }
+
+    [Fact]
+    public void SchedulerIssue_ShredsEveryColumn_Normal()
+    {
+        // Real field capture: process_utilization=3, system_idle=95 -> other = 100-3-95=2;
+        // working_set_delta 77824 bytes -> 77824/1048576 = 0.074218... MB, rounded to 0.07.
+        var r = SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor_normal.xml"));
+
+        Assert.NotNull(r);
+        Assert.Equal(3, r!.SqlCpuUtilization);
+        Assert.Equal(2, r.OtherProcessCpu);
+        Assert.Equal(95, r.SystemIdle);
+        Assert.Equal(100, r.MemoryUtilization);
+        Assert.Equal(99L, r.PageFaults);
+        Assert.Equal(0.07m, r.WorkingSetDeltaMb);
+        Assert.False(SystemHealthSignificance.IsSignificant(r));
+    }
+
+    [Fact]
+    public void SchedulerIssue_OtherProcessCpu_Significant()
+    {
+        // Derived from the real "normal" capture (system_idle changed 95 -> 40) to exercise the
+        // other-process-CPU arm: other = 100-3-40=57 >= 50.
+        var r = SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor_other_process_cpu.xml"));
+        Assert.NotNull(r);
+        Assert.Equal(57, r!.OtherProcessCpu);
+        Assert.True(SystemHealthSignificance.IsSignificant(r));
+    }
+
+    [Fact]
+    public void SchedulerIssue_LowMemory_Significant()
+    {
+        // Derived from the real "normal" capture (memory_utilization changed 100 -> 50) to exercise
+        // the low-memory arm: 50 <= 50.
+        var r = SystemHealthParser.ParseSchedulerIssue(LoadFixture("scheduler_monitor_low_memory.xml"));
+        Assert.NotNull(r);
+        Assert.Equal(50, r!.MemoryUtilization);
+        Assert.True(SystemHealthSignificance.IsSignificant(r));
+    }
+
+    [Theory]
+    [InlineData(90, 0, 100, true)]        // sql cpu boundary: 90 is significant
+    [InlineData(89, 0, 100, false)]       // one under, not significant
+    [InlineData(0, 50, 100, true)]        // other-process boundary: 50 is significant (sql=0,idle=50->other=50)
+    [InlineData(0, 51, 100, false)]       // other-process 49, not significant (idle=51->other=49)
+    [InlineData(0, 100, 50, true)]        // memory boundary: 50 is significant
+    [InlineData(0, 100, 51, false)]       // memory 51, not significant
+    [InlineData(null, null, null, false)] // all null -> not significant (SQL null comparison is false)
+    public void SchedulerIssue_SignificanceBoundaries(int? sqlCpu, int? idle, int? mem, bool expected)
+    {
+        var other = sqlCpu is { } sc && idle is { } id ? 100 - sc - id : (int?)null;
+        var r = new SchedulerIssueRecord { SqlCpuUtilization = sqlCpu, OtherProcessCpu = other, SystemIdle = idle, MemoryUtilization = mem };
+        Assert.Equal(expected, SystemHealthSignificance.IsSignificant(r));
     }
 
     // ── SevereErrors (error_reported) ──
@@ -385,7 +440,7 @@ public void ParseEvents_DispatchesEachEventToItsCategory()
     {
         var events = new (string?, string?)[]
         {
-            (SystemHealthParser.SchedulerMonitorEvent, LoadFixture("scheduler_monitor.xml")),
+            (SystemHealthParser.SchedulerMonitorEvent, LoadFixture("scheduler_monitor_high_sql_cpu.xml")),
             (SystemHealthParser.ErrorReportedEvent, LoadFixture("error_reported.xml")),
             (SystemHealthParser.SpServerDiagnosticsEvent, LoadFixture("sp_server_diagnostics_system.xml")),
             (SystemHealthParser.SpServerDiagnosticsEvent, LoadFixture("sp_server_diagnostics_resource.xml")),
@@ -410,7 +465,7 @@ public void ParseEvents_DispatchesEachEventToItsCategory()
         Assert.Single(result.MemoryConditions);
         Assert.Single(result.CpuTasks);
         Assert.Equal(2, result.IoIssues.Count);
-        Assert.Equal(10, result.SchedulerIssues[0].SchedulerId);
+        Assert.Equal(94, result.SchedulerIssues[0].SqlCpuUtilization);
         Assert.Equal(823, result.SevereErrors[0].ErrorNumber);
         Assert.Equal(6, result.SystemHealth[0].BadPagesDetected);
     }