Update checkpoint from remote nodes replicas (opensearch-project#13888) (opensearch-project#14062) (opensearch-project#14094)

gbbafna · kkewwei · commit e4bfda7f9643 · 2024-07-24T15:40:41.000+08:00
Signed-off-by: Gaurav Bafna &lt;gbbafna@amazon.com&gt;
Signed-off-by: kkewwei &lt;kkewwei@163.com&gt;
diff --git a/server/src/internalClusterTest/java/org/opensearch/remotemigration/MigrationBaseTestCase.java b/server/src/internalClusterTest/java/org/opensearch/remotemigration/MigrationBaseTestCase.java
@@ -186,10 +186,11 @@ private Thread getIndexingThread() {
                     indexSingleDoc(indexName);
                     long currentDocCount = indexedDocs.incrementAndGet();
                     if (currentDocCount > 0 && currentDocCount % refreshFrequency == 0) {
-                        logger.info("--> [iteration {}] flushing index", currentDocCount);
                         if (rarely()) {
+                            logger.info("--> [iteration {}] flushing index", currentDocCount);
                             client().admin().indices().prepareFlush(indexName).get();
                         } else {
+                            logger.info("--> [iteration {}] refreshing index", currentDocCount);
                             client().admin().indices().prepareRefresh(indexName).get();
                         }
                     }
diff --git a/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemoteMigrationIndexMetadataUpdateIT.java b/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemoteMigrationIndexMetadataUpdateIT.java
@@ -273,6 +273,7 @@ initalMetadataVersion < internalCluster().client()
      * After shard relocation completes, shuts down the docrep nodes and asserts remote
      * index settings are applied even when the index is in YELLOW state
      */
+    @AwaitsFix(bugUrl = "https://github.com/opensearch-project/OpenSearch/issues/13737")
     public void testIndexSettingsUpdatedEvenForMisconfiguredReplicas() throws Exception {
         internalCluster().startClusterManagerOnlyNode();
 
@@ -329,6 +330,7 @@ public void testIndexSettingsUpdatedEvenForMisconfiguredReplicas() throws Except
      * After shard relocation completes, restarts the docrep node holding extra replica shard copy
      * and asserts remote index settings are applied as soon as the docrep replica copy is unassigned
      */
+    @AwaitsFix(bugUrl = "https://github.com/opensearch-project/OpenSearch/issues/13871")
     public void testIndexSettingsUpdatedWhenDocrepNodeIsRestarted() throws Exception {
         internalCluster().startClusterManagerOnlyNode();
 
@@ -469,6 +471,7 @@ public void testRemotePathMetadataAddedWithFirstPrimaryMovingToRemote() throws E
      * exclude docrep nodes, assert that remote index path file exists
      * when shards start relocating to the remote nodes.
      */
+    @AwaitsFix(bugUrl = "https://github.com/opensearch-project/OpenSearch/issues/13939")
     public void testRemoteIndexPathFileExistsAfterMigration() throws Exception {
         String docrepClusterManager = internalCluster().startClusterManagerOnlyNode();
 
diff --git a/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemotePrimaryRelocationIT.java b/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemotePrimaryRelocationIT.java
@@ -8,14 +8,11 @@
 
 package org.opensearch.remotemigration;
 
-import org.opensearch.action.DocWriteResponse;
 import org.opensearch.action.admin.cluster.health.ClusterHealthRequest;
 import org.opensearch.action.admin.cluster.health.ClusterHealthResponse;
 import org.opensearch.action.admin.cluster.repositories.get.GetRepositoriesRequest;
 import org.opensearch.action.admin.cluster.repositories.get.GetRepositoriesResponse;
 import org.opensearch.action.admin.cluster.settings.ClusterUpdateSettingsRequest;
-import org.opensearch.action.delete.DeleteResponse;
-import org.opensearch.action.index.IndexResponse;
 import org.opensearch.client.Client;
 import org.opensearch.client.Requests;
 import org.opensearch.cluster.routing.allocation.command.MoveAllocationCommand;
@@ -66,8 +63,8 @@ public void testRemotePrimaryRelocation() throws Exception {
 
         AtomicInteger numAutoGenDocs = new AtomicInteger();
         final AtomicBoolean finished = new AtomicBoolean(false);
-        Thread indexingThread = getIndexingThread(finished, numAutoGenDocs);
-
+        AsyncIndexingService asyncIndexingService = new AsyncIndexingService("test");
+        asyncIndexingService.startIndexing();
         refresh("test");
 
         // add remote node in mixed mode cluster
@@ -141,17 +138,19 @@ public void testRemotePrimaryRelocation() throws Exception {
         logger.info("-->  relocation from remote to remote  complete");
 
         finished.set(true);
-        indexingThread.join();
+        asyncIndexingService.stopIndexing();
         refresh("test");
-        OpenSearchAssertions.assertHitCount(client().prepareSearch("test").setTrackTotalHits(true).get(), numAutoGenDocs.get());
+        OpenSearchAssertions.assertHitCount(
+            client().prepareSearch("test").setTrackTotalHits(true).get(),
+            asyncIndexingService.getIndexedDocs()
+        );
         OpenSearchAssertions.assertHitCount(
             client().prepareSearch("test")
                 .setTrackTotalHits(true)// extra paranoia ;)
                 .setQuery(QueryBuilders.termQuery("auto", true))
                 .get(),
-            numAutoGenDocs.get()
+            asyncIndexingService.getIndexedDocs()
         );
-
     }
 
     public void testMixedModeRelocation_RemoteSeedingFail() throws Exception {
@@ -165,9 +164,8 @@ public void testMixedModeRelocation_RemoteSeedingFail() throws Exception {
         client().admin().indices().prepareCreate("test").setSettings(indexSettings()).setMapping("field", "type=text").get();
         ensureGreen("test");
 
-        AtomicInteger numAutoGenDocs = new AtomicInteger();
-        final AtomicBoolean finished = new AtomicBoolean(false);
-        Thread indexingThread = getIndexingThread(finished, numAutoGenDocs);
+        AsyncIndexingService asyncIndexingService = new AsyncIndexingService("test");
+        asyncIndexingService.startIndexing();
 
         refresh("test");
 
@@ -209,27 +207,11 @@ public void testMixedModeRelocation_RemoteSeedingFail() throws Exception {
         assertEquals(actionGet.getRelocatingShards(), 0);
         assertEquals(docRepNode, primaryNodeName("test"));
 
-        finished.set(true);
-        indexingThread.join();
+        asyncIndexingService.stopIndexing();
         client().admin()
             .cluster()
             .prepareUpdateSettings()
             .setTransientSettings(Settings.builder().put(RecoverySettings.INDICES_INTERNAL_REMOTE_UPLOAD_TIMEOUT.getKey(), (String) null))
             .get();
     }
-
-    private static Thread getIndexingThread(AtomicBoolean finished, AtomicInteger numAutoGenDocs) {
-        Thread indexingThread = new Thread(() -> {
-            while (finished.get() == false && numAutoGenDocs.get() < 10_000) {
-                IndexResponse indexResponse = client().prepareIndex("test").setId("id").setSource("field", "value").get();
-                assertEquals(DocWriteResponse.Result.CREATED, indexResponse.getResult());
-                DeleteResponse deleteResponse = client().prepareDelete("test", "id").get();
-                assertEquals(DocWriteResponse.Result.DELETED, deleteResponse.getResult());
-                client().prepareIndex("test").setSource("auto", true).get();
-                numAutoGenDocs.incrementAndGet();
-            }
-        });
-        indexingThread.start();
-        return indexingThread;
-    }
 }
diff --git a/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemoteReplicaRecoveryIT.java b/server/src/internalClusterTest/java/org/opensearch/remotemigration/RemoteReplicaRecoveryIT.java
@@ -8,32 +8,27 @@
 
 package org.opensearch.remotemigration;
 
-import com.carrotsearch.randomizedtesting.generators.RandomNumbers;
-
-import org.opensearch.action.DocWriteResponse;
 import org.opensearch.action.admin.cluster.health.ClusterHealthResponse;
 import org.opensearch.action.admin.cluster.settings.ClusterUpdateSettingsRequest;
+import org.opensearch.action.admin.indices.replication.SegmentReplicationStatsResponse;
 import org.opensearch.action.admin.indices.settings.put.UpdateSettingsRequest;
-import org.opensearch.action.delete.DeleteResponse;
-import org.opensearch.action.index.IndexResponse;
 import org.opensearch.cluster.metadata.IndexMetadata;
 import org.opensearch.cluster.routing.allocation.command.MoveAllocationCommand;
 import org.opensearch.common.Priority;
 import org.opensearch.common.settings.Settings;
 import org.opensearch.common.unit.TimeValue;
+import org.opensearch.index.SegmentReplicationPerGroupStats;
 import org.opensearch.index.query.QueryBuilders;
 import org.opensearch.test.OpenSearchIntegTestCase;
 import org.opensearch.test.hamcrest.OpenSearchAssertions;
 
-import java.util.concurrent.atomic.AtomicBoolean;
-import java.util.concurrent.atomic.AtomicInteger;
+import java.util.concurrent.TimeUnit;
 
 import static org.opensearch.node.remotestore.RemoteStoreNodeService.MIGRATION_DIRECTION_SETTING;
 import static org.opensearch.node.remotestore.RemoteStoreNodeService.REMOTE_STORE_COMPATIBILITY_MODE_SETTING;
 import static org.opensearch.test.hamcrest.OpenSearchAssertions.assertAcked;
 
 @OpenSearchIntegTestCase.ClusterScope(scope = OpenSearchIntegTestCase.Scope.TEST, numDataNodes = 0, autoManageMasterNodes = false)
-
 public class RemoteReplicaRecoveryIT extends MigrationBaseTestCase {
 
     protected int maximumNumberOfShards() {
@@ -52,6 +47,7 @@ protected int minimumNumberOfReplicas() {
     Brings up new replica copies on remote and docrep nodes, when primary is on a remote node
     Live indexing is happening meanwhile
     */
+    @AwaitsFix(bugUrl = "https://github.com/opensearch-project/OpenSearch/issues/13473")
     public void testReplicaRecovery() throws Exception {
         internalCluster().setBootstrapClusterManagerNodeIndex(0);
         String primaryNode = internalCluster().startNode();
@@ -63,10 +59,8 @@ public void testReplicaRecovery() throws Exception {
         client().admin().indices().prepareCreate("test").setSettings(indexSettings()).setMapping("field", "type=text").get();
         String replicaNode = internalCluster().startNode();
         ensureGreen("test");
-
-        AtomicInteger numAutoGenDocs = new AtomicInteger();
-        final AtomicBoolean finished = new AtomicBoolean(false);
-        Thread indexingThread = getThread(finished, numAutoGenDocs);
+        AsyncIndexingService asyncIndexingService = new AsyncIndexingService("test");
+        asyncIndexingService.startIndexing();
 
         refresh("test");
 
@@ -78,12 +72,10 @@ public void testReplicaRecovery() throws Exception {
         updateSettingsRequest.persistentSettings(Settings.builder().put(MIGRATION_DIRECTION_SETTING.getKey(), "remote_store"));
         assertAcked(client().admin().cluster().updateSettings(updateSettingsRequest).actionGet());
 
-        String remoteNode2 = internalCluster().startNode();
+        internalCluster().startNode();
         internalCluster().validateClusterFormed();
 
         // identify the primary
-
-        Thread.sleep(RandomNumbers.randomIntBetween(random(), 0, 2000));
         logger.info("-->  relocating primary from {} to {} ", primaryNode, remoteNode);
         client().admin()
             .cluster()
@@ -102,7 +94,6 @@ public void testReplicaRecovery() throws Exception {
 
         assertEquals(0, clusterHealthResponse.getRelocatingShards());
         logger.info("-->  relocation of primary from docrep to remote  complete");
-        Thread.sleep(RandomNumbers.randomIntBetween(random(), 0, 2000));
 
         logger.info("--> getting up the new replicas now to doc rep node as well as remote node ");
         // Increase replica count to 3
@@ -129,52 +120,33 @@ public void testReplicaRecovery() throws Exception {
         logger.info("-->  replica  is up now on another docrep now as well as remote node");
 
         assertEquals(0, clusterHealthResponse.getRelocatingShards());
+        asyncIndexingService.stopIndexing();
+        refresh("test");
 
-        Thread.sleep(RandomNumbers.randomIntBetween(random(), 0, 2000));
+        // segrep lag should be zero
+        assertBusy(() -> {
+            SegmentReplicationStatsResponse segmentReplicationStatsResponse = dataNodeClient().admin()
+                .indices()
+                .prepareSegmentReplicationStats("test")
+                .setDetailed(true)
+                .execute()
+                .actionGet();
+            SegmentReplicationPerGroupStats perGroupStats = segmentReplicationStatsResponse.getReplicationStats().get("test").get(0);
+            assertEquals(segmentReplicationStatsResponse.getReplicationStats().size(), 1);
+            perGroupStats.getReplicaStats().stream().forEach(e -> assertEquals(e.getCurrentReplicationLagMillis(), 0));
+        }, 20, TimeUnit.SECONDS);
 
-        // Stop replicas on docrep now.
-        // ToDo : Remove once we have dual replication enabled
-        client().admin()
-            .indices()
-            .updateSettings(
-                new UpdateSettingsRequest("test").settings(
-                    Settings.builder()
-                        .put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 1)
-                        .put("index.routing.allocation.exclude._name", primaryNode + "," + replicaNode)
-                        .build()
-                )
-            )
-            .get();
-
-        finished.set(true);
-        indexingThread.join();
-        refresh("test");
-        OpenSearchAssertions.assertHitCount(client().prepareSearch("test").setTrackTotalHits(true).get(), numAutoGenDocs.get());
+        OpenSearchAssertions.assertHitCount(
+            client().prepareSearch("test").setTrackTotalHits(true).get(),
+            asyncIndexingService.getIndexedDocs()
+        );
         OpenSearchAssertions.assertHitCount(
             client().prepareSearch("test")
                 .setTrackTotalHits(true)// extra paranoia ;)
                 .setQuery(QueryBuilders.termQuery("auto", true))
-                // .setPreference("_prefer_nodes:" + (remoteNode+ "," + remoteNode2))
                 .get(),
-            numAutoGenDocs.get()
+            asyncIndexingService.getIndexedDocs()
         );
 
     }
-
-    private Thread getThread(AtomicBoolean finished, AtomicInteger numAutoGenDocs) {
-        Thread indexingThread = new Thread(() -> {
-            while (finished.get() == false && numAutoGenDocs.get() < 100) {
-                IndexResponse indexResponse = client().prepareIndex("test").setId("id").setSource("field", "value").get();
-                assertEquals(DocWriteResponse.Result.CREATED, indexResponse.getResult());
-                DeleteResponse deleteResponse = client().prepareDelete("test", "id").get();
-                assertEquals(DocWriteResponse.Result.DELETED, deleteResponse.getResult());
-                client().prepareIndex("test").setSource("auto", true).get();
-                numAutoGenDocs.incrementAndGet();
-                logger.info("Indexed {} docs here", numAutoGenDocs.get());
-            }
-        });
-        indexingThread.start();
-        return indexingThread;
-    }
-
 }
diff --git a/server/src/main/java/org/opensearch/index/shard/IndexShard.java b/server/src/main/java/org/opensearch/index/shard/IndexShard.java
@@ -525,7 +525,6 @@ public boolean shouldSeedRemoteStore() {
     public Function<String, Boolean> isShardOnRemoteEnabledNode = nodeId -> {
         DiscoveryNode node = discoveryNodes.get(nodeId);
         if (node != null) {
-            logger.trace("Node {} has remote_enabled as {}", nodeId, node.isRemoteStoreNode());
             return node.isRemoteStoreNode();
         }
         return false;
diff --git a/server/src/main/java/org/opensearch/indices/replication/SegmentReplicationTargetService.java b/server/src/main/java/org/opensearch/indices/replication/SegmentReplicationTargetService.java
@@ -384,7 +384,7 @@ private void logReplicationFailure(SegmentReplicationState state, ReplicationFai
     protected void updateVisibleCheckpoint(long replicationId, IndexShard replicaShard) {
         // Update replication checkpoint on source via transport call only supported for remote store integration. For node-
         // node communication, checkpoint update is piggy-backed to GET_SEGMENT_FILES transport call
-        if (replicaShard.indexSettings().isRemoteStoreEnabled() == false) {
+        if (replicaShard.indexSettings().isAssignedOnRemoteNode() == false) {
             return;
         }
         ShardRouting primaryShard = clusterService.state().routingTable().shardRoutingTable(replicaShard.shardId()).primaryShard();
diff --git a/server/src/main/java/org/opensearch/indices/replication/common/ReplicationTarget.java b/server/src/main/java/org/opensearch/indices/replication/common/ReplicationTarget.java
@@ -91,7 +91,7 @@ public ReplicationTarget(String name, IndexShard indexShard, ReplicationLuceneIn
         // make sure the store is not released until we are done.
         this.cancellableThreads = new CancellableThreads();
         store.incRef();
-        if (indexShard.indexSettings().isRemoteStoreEnabled()) {
+        if (indexShard.indexSettings().isAssignedOnRemoteNode()) {
             indexShard.remoteStore().incRef();
         }
     }
@@ -284,7 +284,7 @@ protected void closeInternal() {
         try {
             store.decRef();
         } finally {
-            if (indexShard.indexSettings().isRemoteStoreEnabled()) {
+            if (indexShard.indexSettings().isAssignedOnRemoteNode()) {
                 indexShard.remoteStore().decRef();
             }
         }

Original file line number	Diff line number	Diff line change
`@@ -186,10 +186,11 @@ private Thread getIndexingThread() {`
`186`	`186`	`indexSingleDoc(indexName);`
`187`	`187`	`long currentDocCount = indexedDocs.incrementAndGet();`
`188`	`188`	`if (currentDocCount > 0 && currentDocCount % refreshFrequency == 0) {`
`189`		`- logger.info("--> [iteration {}] flushing index", currentDocCount);`
`190`	`189`	`if (rarely()) {`
	`190`	`+ logger.info("--> [iteration {}] flushing index", currentDocCount);`
`191`	`191`	`client().admin().indices().prepareFlush(indexName).get();`
`192`	`192`	`} else {`
	`193`	`+ logger.info("--> [iteration {}] refreshing index", currentDocCount);`
`193`	`194`	`client().admin().indices().prepareRefresh(indexName).get();`
`194`	`195`	`}`
`195`	`196`	`}`
Original file line number	Diff line number	Diff line change
`@@ -525,7 +525,6 @@ public boolean shouldSeedRemoteStore() {`
`525`	`525`	`public Function<String, Boolean> isShardOnRemoteEnabledNode = nodeId -> {`
`526`	`526`	`DiscoveryNode node = discoveryNodes.get(nodeId);`
`527`	`527`	`if (node != null) {`
`528`		`- logger.trace("Node {} has remote_enabled as {}", nodeId, node.isRemoteStoreNode());`
`529`	`528`	`return node.isRemoteStoreNode();`
`530`	`529`	`}`
`531`	`530`	`return false;`
Original file line number	Diff line number	Diff line change
`@@ -384,7 +384,7 @@ private void logReplicationFailure(SegmentReplicationState state, ReplicationFai`
`384`	`384`	`protected void updateVisibleCheckpoint(long replicationId, IndexShard replicaShard) {`
`385`	`385`	`// Update replication checkpoint on source via transport call only supported for remote store integration. For node-`
`386`	`386`	`// node communication, checkpoint update is piggy-backed to GET_SEGMENT_FILES transport call`
`387`		`- if (replicaShard.indexSettings().isRemoteStoreEnabled() == false) {`
	`387`	`+ if (replicaShard.indexSettings().isAssignedOnRemoteNode() == false) {`
`388`	`388`	`return;`
`389`	`389`	`}`
`390`	`390`	`ShardRouting primaryShard = clusterService.state().routingTable().shardRoutingTable(replicaShard.shardId()).primaryShard();`
Original file line number	Diff line number	Diff line change
`@@ -91,7 +91,7 @@ public ReplicationTarget(String name, IndexShard indexShard, ReplicationLuceneIn`
`91`	`91`	`// make sure the store is not released until we are done.`
`92`	`92`	`this.cancellableThreads = new CancellableThreads();`
`93`	`93`	`store.incRef();`
`94`		`- if (indexShard.indexSettings().isRemoteStoreEnabled()) {`
	`94`	`+ if (indexShard.indexSettings().isAssignedOnRemoteNode()) {`
`95`	`95`	`indexShard.remoteStore().incRef();`
`96`	`96`	`}`
`97`	`97`	`}`
`@@ -284,7 +284,7 @@ protected void closeInternal() {`
`284`	`284`	`try {`
`285`	`285`	`store.decRef();`
`286`	`286`	`} finally {`
`287`		`- if (indexShard.indexSettings().isRemoteStoreEnabled()) {`
	`287`	`+ if (indexShard.indexSettings().isAssignedOnRemoteNode()) {`
`288`	`288`	`indexShard.remoteStore().decRef();`
`289`	`289`	`}`
`290`	`290`	`}`