Tracing for deep search path

Signed-off-by: David Zane <[email protected]>

CHANGELOG.md

@@ -16,6 +16,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Allow to pass the list settings through environment variables (like [], ["a", "b", "c"], ...) ([#10625](https://github.com/opensearch-project/OpenSearch/pull/10625))
 - [Admission Control] Integrate CPU AC with ResourceUsageCollector and add CPU AC stats to nodes/stats ([#10887](https://github.com/opensearch-project/OpenSearch/pull/10887))
 - [S3 Repository] Add setting to control connection count for sync client ([#12028](https://github.com/opensearch-project/OpenSearch/pull/12028))
+- Tracing for deep search path ([#12103](https://github.com/opensearch-project/OpenSearch/pull/12103))
 
 ### Dependencies
 - Bump `log4j-core` from 2.18.0 to 2.19.0

libs/telemetry/src/main/java/org/opensearch/telemetry/tracing/DefaultTracer.java

@@ -24,7 +24,7 @@
  *  @opensearch.internal
  */
 @InternalApi
-class DefaultTracer implements Tracer {
+public class DefaultTracer implements Tracer {
     /**
      * Current thread name.
      */

server/src/main/java/org/opensearch/action/search/AbstractSearchAsyncAction.java

@@ -58,6 +58,7 @@
 import org.opensearch.search.internal.SearchContext;
 import org.opensearch.search.internal.ShardSearchRequest;
 import org.opensearch.search.pipeline.PipelinedRequest;
+import org.opensearch.telemetry.tracing.SpanScope;
 import org.opensearch.transport.Transport;
 
 import java.util.ArrayDeque;
@@ -220,6 +221,7 @@ public final void start() {
                     null
                 )
             );
+            onRequestEnd(searchRequestContext);
             return;
         }
         executePhase(this);
@@ -439,21 +441,23 @@ private void onPhaseEnd(SearchRequestContext searchRequestContext) {
         }
     }
 
-    void onPhaseStart(SearchPhase phase) {
+    void onPhaseStart(SearchPhase phase, SearchRequestContext searchRequestContext) {
         setCurrentPhase(phase);
         if (SearchPhaseName.isValidName(phase.getName())) {
-            this.searchRequestContext.getSearchRequestOperationsListener().onPhaseStart(this);
+            this.searchRequestContext.getSearchRequestOperationsListener().onPhaseStart(this, searchRequestContext);
         }
     }
 
     private void onRequestEnd(SearchRequestContext searchRequestContext) {
-        this.searchRequestContext.getSearchRequestOperationsListener().onRequestEnd(this, searchRequestContext);
+        this.searchRequestContext.getSearchRequestOperationsListener().onRequestEnd(searchRequestContext);
     }
 
     private void executePhase(SearchPhase phase) {
         try {
-            onPhaseStart(phase);
-            phase.recordAndRun();
+            onPhaseStart(phase, searchRequestContext);
+            try (SpanScope spanScope = searchRequestContext.getTracer().withSpanInScope(searchRequestContext.getPhaseSpan())) {
+                phase.recordAndRun();
+            }
         } catch (Exception e) {
             if (logger.isDebugEnabled()) {
                 logger.debug(new ParameterizedMessage("Failed to execute [{}] while moving to [{}] phase", request, phase.getName()), e);
@@ -705,6 +709,7 @@ public void sendSearchResponse(InternalSearchResponse internalSearchResponse, At
                     searchContextId = null;
                 }
             }
+            searchRequestContext.setSearchTask(getTask());
             searchRequestContext.setTotalHits(internalSearchResponse.hits().getTotalHits());
             searchRequestContext.setShardStats(results.getNumShards(), successfulOps.get(), skippedOps.get(), failures.length);
             onPhaseEnd(searchRequestContext);
@@ -717,7 +722,7 @@ public void sendSearchResponse(InternalSearchResponse internalSearchResponse, At
     @Override
     public final void onPhaseFailure(SearchPhase phase, String msg, Throwable cause) {
         if (SearchPhaseName.isValidName(phase.getName())) {
-            this.searchRequestContext.getSearchRequestOperationsListener().onPhaseFailure(this);
+            this.searchRequestContext.getSearchRequestOperationsListener().onPhaseFailure(this, searchRequestContext);
         }
         raisePhaseFailure(new SearchPhaseExecutionException(phase.getName(), msg, cause, buildShardFailures()));
     }

server/src/main/java/org/opensearch/action/search/SearchRequestContext.java

@@ -10,6 +10,10 @@
 
 import org.apache.lucene.search.TotalHits;
 import org.opensearch.common.annotation.InternalApi;
+import org.opensearch.telemetry.tracing.Span;
+import org.opensearch.telemetry.tracing.Tracer;
+import org.opensearch.telemetry.tracing.noop.NoopSpan;
+import org.opensearch.telemetry.tracing.noop.NoopTracer;
 
 import java.util.EnumMap;
 import java.util.HashMap;
@@ -23,20 +27,45 @@
  */
 @InternalApi
 class SearchRequestContext {
+    private final SearchRequest searchRequest;
+    private SearchTask searchTask;
     private final SearchRequestOperationsListener searchRequestOperationsListener;
     private long absoluteStartNanos;
     private final Map<String, Long> phaseTookMap;
     private TotalHits totalHits;
     private final EnumMap<ShardStatsFieldNames, Integer> shardStats;
+    private final Tracer tracer;
+    private Span requestSpan;
+    private Span phaseSpan;
 
-    private final SearchRequest searchRequest;
+    /**
+     * This constructor is for testing only
+     */
+    SearchRequestContext(SearchRequestOperationsListener searchRequestOperationsListener, SearchRequest searchRequest) {
+        this(searchRequestOperationsListener, searchRequest, NoopTracer.INSTANCE);
+    }
 
-    SearchRequestContext(final SearchRequestOperationsListener searchRequestOperationsListener, final SearchRequest searchRequest) {
+    SearchRequestContext(SearchRequestOperationsListener searchRequestOperationsListener, SearchRequest searchRequest, Tracer tracer) {
         this.searchRequestOperationsListener = searchRequestOperationsListener;
+        this.searchRequest = searchRequest;
+        this.tracer = tracer;
         this.absoluteStartNanos = System.nanoTime();
         this.phaseTookMap = new HashMap<>();
         this.shardStats = new EnumMap<>(ShardStatsFieldNames.class);
-        this.searchRequest = searchRequest;
+        this.requestSpan = NoopSpan.INSTANCE;
+        this.phaseSpan = NoopSpan.INSTANCE;
+    }
+
+    SearchRequest getSearchRequest() {
+        return searchRequest;
+    }
+
+    void setSearchTask(SearchTask searchTask) {
+        this.searchTask = searchTask;
+    }
+
+    SearchTask getSearchTask() {
+        return searchTask;
     }
 
     SearchRequestOperationsListener getSearchRequestOperationsListener() {
@@ -107,6 +136,26 @@ String formattedShardStats() {
             );
         }
     }
+
+    Tracer getTracer() {
+        return tracer;
+    }
+
+    void setRequestSpan(Span requestSpan) {
+        this.requestSpan = requestSpan;
+    }
+
+    Span getRequestSpan() {
+        return requestSpan;
+    }
+
+    void setPhaseSpan(Span phaseSpan) {
+        this.phaseSpan = phaseSpan;
+    }
+
+    Span getPhaseSpan() {
+        return phaseSpan;
+    }
 }
 
 enum ShardStatsFieldNames {

server/src/main/java/org/opensearch/action/search/SearchRequestCoordinatorTrace.java

@@ -0,0 +1,71 @@
+/*
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * The OpenSearch Contributors require contributions made to
+ * this file be licensed under the Apache-2.0 license or a
+ * compatible open source license.
+ */
+
+package org.opensearch.action.search;
+
+import org.opensearch.telemetry.tracing.AttributeNames;
+import org.opensearch.telemetry.tracing.Span;
+import org.opensearch.telemetry.tracing.SpanBuilder;
+import org.opensearch.telemetry.tracing.SpanContext;
+import org.opensearch.telemetry.tracing.Tracer;
+
+import static org.opensearch.core.common.Strings.capitalize;
+
+/**
+ * Listener for search request tracing on the coordinator node
+ *
+ * @opensearch.internal
+ */
+public final class SearchRequestCoordinatorTrace extends SearchRequestOperationsListener {
+    private final Tracer tracer;
+
+    public SearchRequestCoordinatorTrace(Tracer tracer) {
+        this.tracer = tracer;
+    }
+
+    @Override
+    void onPhaseStart(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
+        searchRequestContext.setPhaseSpan(
+            tracer.startSpan(
+                SpanBuilder.from(
+                    "coord" + capitalize(context.getCurrentPhase().getName()),
+                    new SpanContext(searchRequestContext.getRequestSpan())
+                )
+            )
+        );
+    }
+
+    @Override
+    void onPhaseEnd(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
+        searchRequestContext.getPhaseSpan().endSpan();
+    }
+
+    @Override
+    void onPhaseFailure(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
+        searchRequestContext.getPhaseSpan().endSpan();
+    }
+
+    @Override
+    void onRequestEnd(SearchRequestContext searchRequestContext) {
+        Span requestSpan = searchRequestContext.getRequestSpan();
+
+        // add response-related attributes on request end
+        requestSpan.addAttribute(
+            AttributeNames.TOTAL_HITS,
+            searchRequestContext.totalHits() == null ? "0" : searchRequestContext.totalHits().toString()
+        );
+        requestSpan.addAttribute(
+            AttributeNames.SHARDS,
+            searchRequestContext.formattedShardStats().isEmpty() ? "null" : searchRequestContext.formattedShardStats()
+        );
+        requestSpan.addAttribute(
+            AttributeNames.SOURCE,
+            searchRequestContext.getSearchRequest().source() == null ? "null" : searchRequestContext.getSearchRequest().source().toString()
+        );
+    }
+}

server/src/main/java/org/opensearch/action/search/SearchRequestOperationsListener.java

@@ -31,15 +31,15 @@ protected SearchRequestOperationsListener(final boolean enabled) {
         this.enabled = enabled;
     }
 
-    abstract void onPhaseStart(SearchPhaseContext context);
+    abstract void onPhaseStart(SearchPhaseContext context, SearchRequestContext searchRequestContext);
 
     abstract void onPhaseEnd(SearchPhaseContext context, SearchRequestContext searchRequestContext);
 
-    abstract void onPhaseFailure(SearchPhaseContext context);
+    abstract void onPhaseFailure(SearchPhaseContext context, SearchRequestContext searchRequestContext);
 
     void onRequestStart(SearchRequestContext searchRequestContext) {}
 
-    void onRequestEnd(SearchPhaseContext context, SearchRequestContext searchRequestContext) {}
+    void onRequestEnd(SearchRequestContext searchRequestContext) {}
 
     boolean isEnabled(SearchRequest searchRequest) {
         return isEnabled();
@@ -69,10 +69,10 @@ static final class CompositeListener extends SearchRequestOperationsListener {
         }
 
         @Override
-        void onPhaseStart(SearchPhaseContext context) {
+        void onPhaseStart(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
             for (SearchRequestOperationsListener listener : listeners) {
                 try {
-                    listener.onPhaseStart(context);
+                    listener.onPhaseStart(context, searchRequestContext);
                 } catch (Exception e) {
                     logger.warn(() -> new ParameterizedMessage("onPhaseStart listener [{}] failed", listener), e);
                 }
@@ -91,10 +91,10 @@ void onPhaseEnd(SearchPhaseContext context, SearchRequestContext searchRequestCo
         }
 
         @Override
-        void onPhaseFailure(SearchPhaseContext context) {
+        void onPhaseFailure(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
             for (SearchRequestOperationsListener listener : listeners) {
                 try {
-                    listener.onPhaseFailure(context);
+                    listener.onPhaseFailure(context, searchRequestContext);
                 } catch (Exception e) {
                     logger.warn(() -> new ParameterizedMessage("onPhaseFailure listener [{}] failed", listener), e);
                 }
@@ -113,10 +113,10 @@ void onRequestStart(SearchRequestContext searchRequestContext) {
         }
 
         @Override
-        public void onRequestEnd(SearchPhaseContext context, SearchRequestContext searchRequestContext) {
+        public void onRequestEnd(SearchRequestContext searchRequestContext) {
             for (SearchRequestOperationsListener listener : listeners) {
                 try {
-                    listener.onRequestEnd(context, searchRequestContext);
+                    listener.onRequestEnd(searchRequestContext);
                 } catch (Exception e) {
                     logger.warn(() -> new ParameterizedMessage("onRequestEnd listener [{}] failed", listener), e);
                 }