Support resetting IndexShard engine on the fly

elastic · Jan 17, 2025 · 588aa8d · 588aa8d
1 parent a867127
commit 588aa8d
Show file tree

Hide file tree

Showing 2 changed files with 52 additions and 0 deletions.
diff --git a/server/src/main/java/org/elasticsearch/index/shard/IndexShard.java b/server/src/main/java/org/elasticsearch/index/shard/IndexShard.java
@@ -4361,6 +4361,31 @@ public void afterRefresh(boolean didRefresh) {
         }
     }
 
+    /**
+     * Reset the current engine to a new one without doing translog recovery.
+     */
+    public void resetEngine() throws IOException {
+        assert Thread.holdsLock(mutex) == false : "resetting engine under mutex";
+        assert getActiveOperationsCount() == OPERATIONS_BLOCKED
+            : "resetting engine without blocking operations; active operations are [" + getActiveOperationsCount() + ']';
+        var engineConfig = newEngineConfig(replicationTracker);
+        try {
+            synchronized (engineMutex) {
+                verifyNotClosed();
+                IOUtils.close(currentEngineReference.get());
+
+                var newEngine = createEngine(engineConfig);
+                currentEngineReference.set(newEngine);
+                onNewEngine(newEngine);
+                active.set(true);
+            }
+            onSettingsChanged();
+            checkAndCallWaitForEngineOrClosedShardListeners();
+        } catch (Exception e) {
+            failShard("Unable to reset engine", e);
+        }
+    }
+
     /**
      * Rollback the current engine to the safe commit, then replay local translog up to the global checkpoint.
      */

diff --git a/server/src/test/java/org/elasticsearch/index/shard/IndexShardTests.java b/server/src/test/java/org/elasticsearch/index/shard/IndexShardTests.java
@@ -82,6 +82,7 @@
 import org.elasticsearch.index.engine.EngineTestCase;
 import org.elasticsearch.index.engine.InternalEngine;
 import org.elasticsearch.index.engine.InternalEngineFactory;
+import org.elasticsearch.index.engine.NoOpEngine;
 import org.elasticsearch.index.engine.ReadOnlyEngine;
 import org.elasticsearch.index.engine.Segment;
 import org.elasticsearch.index.engine.VersionConflictEngineException;
@@ -4574,6 +4575,32 @@ public void testResetEngine() throws Exception {
         closeShard(shard, false);
     }
 
+    public void testResetEngineToNoOpEngine() throws Exception {
+        var newEngineCreated = new CountDownLatch(2);
+        var indexShard = newStartedShard(true, Settings.EMPTY, config -> {
+            try {
+                return new NoOpEngine(config);
+            } finally {
+                newEngineCreated.countDown();
+            }
+        });
+        var newEngineNotification = new CountDownLatch(1);
+        indexShard.waitForEngineOrClosedShard(ActionListener.running(newEngineNotification::countDown));
+
+        var onAcquired = new PlainActionFuture<Releasable>();
+        indexShard.acquireAllPrimaryOperationsPermits(onAcquired, TimeValue.timeValueMinutes(1L));
+        try (var permit = safeGet(onAcquired)) {
+            indexShard.resetEngine();
+        }
+        safeAwait(newEngineCreated);
+        safeAwait(newEngineNotification);
+
+        assertThat(indexShard.getEngine(), instanceOf(NoOpEngine.class));
+        assertTrue(indexShard.isActive());
+
+        closeShard(indexShard, false);
+    }
+
     /**
      * This test simulates a scenario seen rarely in ConcurrentSeqNoVersioningIT. Closing a shard while engine is inside
      * resetEngineToGlobalCheckpoint can lead to check index failure in integration tests.