server(metrics): alarm provider mode gauge + provider switch counter (Task 13)

src/ZB.MOM.WW.MxGateway.Server/Metrics/GatewayMetrics.cs

@@ -1,5 +1,6 @@
 using System.Collections.Concurrent;
 using System.Diagnostics.Metrics;
+using System.Globalization;
 
 namespace ZB.MOM.WW.MxGateway.Server.Metrics;
 
@@ -22,6 +23,7 @@ public sealed class GatewayMetrics : IDisposable
     private readonly Counter<long> _heartbeatFailuresCounter;
     private readonly Counter<long> _streamDisconnectsCounter;
     private readonly Counter<long> _retryAttemptsCounter;
+    private readonly Counter<long> _alarmProviderSwitchesCounter;
     private readonly Histogram<double> _workerStartupLatencyHistogram;
     private readonly Histogram<double> _commandLatencyHistogram;
     private readonly Histogram<double> _eventStreamSendLatencyHistogram;
@@ -34,6 +36,7 @@ public sealed class GatewayMetrics : IDisposable
     private int _workersRunning;
     private int _workerEventQueueDepth;
     private int _grpcEventStreamQueueDepth;
+    private int _alarmProviderMode;
     private long _sessionsOpened;
     private long _sessionsClosed;
     private long _commandsStarted;
@@ -68,6 +71,7 @@ public sealed class GatewayMetrics : IDisposable
         _heartbeatFailuresCounter = _meter.CreateCounter<long>("mxgateway.heartbeats.failed");
         _streamDisconnectsCounter = _meter.CreateCounter<long>("mxgateway.grpc.streams.disconnected");
         _retryAttemptsCounter = _meter.CreateCounter<long>("mxgateway.retries.attempted");
+        _alarmProviderSwitchesCounter = _meter.CreateCounter<long>("mxgateway.alarms.provider_switches");
         _workerStartupLatencyHistogram = _meter.CreateHistogram<double>("mxgateway.workers.startup.duration", "s");
         _commandLatencyHistogram = _meter.CreateHistogram<double>("mxgateway.commands.duration", "s");
         _eventStreamSendLatencyHistogram = _meter.CreateHistogram<double>("mxgateway.events.stream_send.duration", "s");
@@ -76,6 +80,7 @@ public sealed class GatewayMetrics : IDisposable
         _meter.CreateObservableGauge("mxgateway.workers.running", GetWorkersRunning);
         _meter.CreateObservableGauge("mxgateway.events.worker_queue.depth", GetWorkerEventQueueDepth);
         _meter.CreateObservableGauge("mxgateway.events.grpc_stream_queue.depth", GetGrpcEventStreamQueueDepth);
+        _meter.CreateObservableGauge("mxgateway.alarms.provider_mode", GetAlarmProviderMode);
     }
 
     /// <summary>
@@ -377,6 +382,26 @@ public sealed class GatewayMetrics : IDisposable
         _retryAttemptsCounter.Add(1, new KeyValuePair<string, object?>("area", area));
     }
 
+    /// <summary>
+    /// Records that the alarm provider switched modes and updates the current provider mode gauge.
+    /// </summary>
+    /// <param name="fromMode">Provider mode before the switch (1=alarmmgr, 2=subtag, 0=unknown).</param>
+    /// <param name="toMode">Provider mode after the switch (1=alarmmgr, 2=subtag, 0=unknown).</param>
+    /// <param name="reason">Human-readable reason for the switch.</param>
+    public void AlarmProviderSwitched(int fromMode, int toMode, string reason)
+    {
+        lock (_syncRoot)
+        {
+            _alarmProviderMode = toMode;
+        }
+
+        _alarmProviderSwitchesCounter.Add(
+            1,
+            new KeyValuePair<string, object?>("from", fromMode.ToString(CultureInfo.InvariantCulture)),
+            new KeyValuePair<string, object?>("to", toMode.ToString(CultureInfo.InvariantCulture)),
+            new KeyValuePair<string, object?>("reason", reason ?? string.Empty));
+    }
+
     /// <summary>
     /// Returns a snapshot of all current metric values.
     /// </summary>
@@ -455,6 +480,14 @@ public sealed class GatewayMetrics : IDisposable
         }
     }
 
+    private int GetAlarmProviderMode()
+    {
+        lock (_syncRoot)
+        {
+            return _alarmProviderMode;
+        }
+    }
+
     private static void Increment(Dictionary<string, long> values, string key)
     {
         values.TryGetValue(key, out long currentValue);