Hybrid Search: Adding fixes for schedulingStopWatch start/stop for a race condition (#46594)

* Adding fixes for stopWatch when running hybrid search queries

* Updating changelog

Author: aayush3011

sdk/cosmos/azure-cosmos/CHANGELOG.md

@@ -12,7 +12,7 @@
 * Fixed Null Pointer Exception for query when container recreated with same name. - [PR 45930](https://github.com/Azure/azure-sdk-for-java/pull/45930)
 * Fixed Null Pointer Exception for readMany when container recreated with same name. - [PR 45930](https://github.com/Azure/azure-sdk-for-java/pull/45930)
 * Fixed parameterized query failures for Hybrid Search queries. - [PR 46446](https://github.com/Azure/azure-sdk-for-java/pull/46446)
-* Fixed a rare race in parallel Hybrid Search queries by making internal SchedulingStopwatch start/stop atomic and idempotent. - [PR 46485](https://github.com/Azure/azure-sdk-for-java/pull/46485)
+* Fixed a rare race in parallel Hybrid Search queries by making internal SchedulingStopwatch start/stop atomic and idempotent. - [PR 46594](https://github.com/Azure/azure-sdk-for-java/pull/46594)
 * Fixed the Hybrid Search SchedulingStopWatch to use Cumulative Timing Across all parallel queries. - [PR 46591](https://github.com/Azure/azure-sdk-for-java/pull/46591)
 * Fixed Strong Consistency violation when a single replica in a partition returns a 410 `Lease Not Found`. - [PR 46433](https://github.com/Azure/azure-sdk-for-java/pull/46433)
 

sdk/cosmos/azure-cosmos/src/main/java/com/azure/cosmos/implementation/query/metrics/SchedulingStopwatch.java

@@ -26,28 +26,34 @@ public SchedulingTimeSpan getElapsedTime() {
 
     /**
      * Tells the SchedulingStopwatch know that the process is in a state where it is ready to be worked on,
-     * which in turn starts the stopwatch for for response time and turnaround time.
+     * which in turn starts the stopwatch for response time and turnaround time.
      */
     public void ready() {
         startStopWatch(this.turnaroundTimeStopwatch);
         startStopWatch(this.responseTimeStopwatch);
     }
 
     public void start() {
-        if (!this.runTimeStopwatch.isStarted()) {
+        synchronized (this.runTimeStopwatch) {
+            if (this.runTimeStopwatch.isStarted()) {
+                return;
+            }
             if (!this.responded) {
                 // This is the first time the process got a response, so the response time stopwatch needs to stop.
-                this.responseTimeStopwatch.stop();
+                stopStopWatch(this.responseTimeStopwatch);
                 this.responded = true;
             }
             this.runTimeStopwatch.reset();
-            startStopWatch(this.runTimeStopwatch);
+            this.runTimeStopwatch.start();
         }
     }
 
     public void stop() {
-        if (this.runTimeStopwatch.isStarted()) {
-            stopStopWatch(this.runTimeStopwatch);
+        synchronized (this.runTimeStopwatch) {
+            if (!this.runTimeStopwatch.isStarted()) {
+                return;
+            }
+            this.runTimeStopwatch.stop();
             this.numPreemptions++;
         }
     }
@@ -59,12 +65,18 @@ public void terminate() {
 
     private void startStopWatch(StopWatch stopwatch) {
         synchronized (stopwatch) {
+            if (stopwatch.isStarted()) {
+                return; // idempotent start
+            }
             stopwatch.start();
         }
     }
 
     private void stopStopWatch(StopWatch stopwatch) {
         synchronized (stopwatch) {
+            if (!stopwatch.isStarted()) {
+                return; // idempotent stop
+            }
             stopwatch.stop();
         }
     }