Add Integ Tests

Signed-off-by: Ashish Singh <ssashish@amazon.com>
opensearch-project · Oct 7, 2023 · 4657a37 · 4657a37
1 parent 2b17e06
commit 4657a37
Show file tree

Hide file tree

Showing 3 changed files with 75 additions and 3 deletions.
diff --git a/...rTest/java/org/opensearch/remotestore/AbstractRemoteStoreMockRepositoryIntegTestCase.java b/...rTest/java/org/opensearch/remotestore/AbstractRemoteStoreMockRepositoryIntegTestCase.java
@@ -160,7 +160,7 @@ private String getLocalSegmentFilename(String remoteFilename) {
         return remoteFilename.split(RemoteSegmentStoreDirectory.SEGMENT_NAME_UUID_SEPARATOR)[0];
     }
 
-    private IndexResponse indexSingleDoc() {
+    protected IndexResponse indexSingleDoc() {
         return client().prepareIndex(INDEX_NAME)
             .setId(UUIDs.randomBase64UUID())
             .setSource(randomAlphaOfLength(5), randomAlphaOfLength(5))

diff --git a/...emotestore/RemoteStoreBackpressureIT.java → ...moteStoreBackpressureAndResiliencyIT.java b/...emotestore/RemoteStoreBackpressureIT.java → ...moteStoreBackpressureAndResiliencyIT.java
@@ -12,12 +12,18 @@
 import org.opensearch.action.admin.cluster.remotestore.stats.RemoteStoreStatsResponse;
 import org.opensearch.action.admin.cluster.settings.ClusterUpdateSettingsResponse;
 import org.opensearch.common.settings.Settings;
+import org.opensearch.common.unit.TimeValue;
+import org.opensearch.common.util.concurrent.AbstractAsyncTask;
+import org.opensearch.common.util.concurrent.UncategorizedExecutionException;
 import org.opensearch.core.common.bytes.BytesArray;
 import org.opensearch.core.common.bytes.BytesReference;
 import org.opensearch.core.common.unit.ByteSizeUnit;
 import org.opensearch.core.concurrency.OpenSearchRejectedExecutionException;
 import org.opensearch.core.xcontent.MediaTypeRegistry;
+import org.opensearch.index.IndexService;
 import org.opensearch.index.remote.RemoteSegmentTransferTracker;
+import org.opensearch.index.shard.IndexShard;
+import org.opensearch.indices.IndicesService;
 import org.opensearch.repositories.RepositoriesService;
 import org.opensearch.snapshots.mockstore.MockRepository;
 import org.opensearch.test.OpenSearchIntegTestCase;
@@ -33,7 +39,7 @@
 import static org.opensearch.index.remote.RemoteStorePressureSettings.REMOTE_REFRESH_SEGMENT_PRESSURE_ENABLED;
 
 @OpenSearchIntegTestCase.ClusterScope(scope = OpenSearchIntegTestCase.Scope.TEST, numDataNodes = 0)
-public class RemoteStoreBackpressureIT extends AbstractRemoteStoreMockRepositoryIntegTestCase {
+public class RemoteStoreBackpressureAndResiliencyIT extends AbstractRemoteStoreMockRepositoryIntegTestCase {
     public void testWritesRejectedDueToConsecutiveFailureBreach() throws Exception {
         // Here the doc size of the request remains same throughout the test. After initial indexing, all remote store interactions
         // fail leading to consecutive failure limit getting exceeded and leading to rejections.
@@ -156,4 +162,70 @@ private String generateString(int sizeInBytes) {
         sb.append("}");
         return sb.toString();
     }
+
+    /**
+     * Fixes <a href="https://github.com/opensearch-project/OpenSearch/issues/10398">Github#10398</a>
+     */
+    public void testAsyncTrimTaskSucceeds() {
+        Path location = randomRepoPath().toAbsolutePath();
+        String dataNodeName = setup(location, 0d, "metadata", Long.MAX_VALUE);
+
+        logger.info("Increasing the frequency of async trim task to ensure it runs in background while indexing");
+        IndexService indexService = internalCluster().getInstance(IndicesService.class, dataNodeName).iterator().next();
+        ((AbstractAsyncTask) indexService.getTrimTranslogTask()).setInterval(TimeValue.timeValueMillis(100));
+
+        logger.info("--> Indexing data");
+        indexData(randomIntBetween(2, 5), true);
+        logger.info("--> Indexing succeeded");
+
+        MockRepository translogRepo = (MockRepository) internalCluster().getInstance(RepositoriesService.class, dataNodeName)
+            .repository(TRANSLOG_REPOSITORY_NAME);
+        logger.info("--> Failing all remote store interaction");
+        translogRepo.setRandomControlIOExceptionRate(1d);
+
+        for (int i = 0; i < randomIntBetween(5, 10); i++) {
+            UncategorizedExecutionException exception = assertThrows(UncategorizedExecutionException.class, this::indexSingleDoc);
+            assertEquals("Failed execution", exception.getMessage());
+        }
+
+        translogRepo.setRandomControlIOExceptionRate(0d);
+        indexSingleDoc();
+        logger.info("Indexed single doc successfully");
+    }
+
+    /**
+     * Fixes <a href="https://github.com/opensearch-project/OpenSearch/issues/10400">Github#10400</a>
+     */
+    public void testSkipLoadGlobalCheckpointToReplicationTracker() {
+        Path location = randomRepoPath().toAbsolutePath();
+        String dataNodeName = setup(location, 0d, "metadata", Long.MAX_VALUE);
+
+        logger.info("--> Indexing data");
+        indexData(randomIntBetween(1, 2), true);
+        logger.info("--> Indexing succeeded");
+
+        IndexService indexService = internalCluster().getInstance(IndicesService.class, dataNodeName).iterator().next();
+        IndexShard indexShard = indexService.getShard(0);
+        indexShard.failShard("failing shard", null);
+
+        ensureRed(INDEX_NAME);
+
+        MockRepository translogRepo = (MockRepository) internalCluster().getInstance(RepositoriesService.class, dataNodeName)
+            .repository(TRANSLOG_REPOSITORY_NAME);
+        logger.info("--> Failing all remote store interaction");
+        translogRepo.setRandomControlIOExceptionRate(1d);
+        client().admin().cluster().prepareReroute().setRetryFailed(true).get();
+        // CLuster stays red still as the remote interactions are still failing
+        ensureRed(INDEX_NAME);
+
+        logger.info("Retrying to allocate failed shards");
+        client().admin().cluster().prepareReroute().setRetryFailed(true).get();
+        // CLuster stays red still as the remote interactions are still failing
+        ensureRed(INDEX_NAME);
+
+        logger.info("Stop failing all remote store interactions");
+        translogRepo.setRandomControlIOExceptionRate(0d);
+        client().admin().cluster().prepareReroute().setRetryFailed(true).get();
+        ensureGreen(INDEX_NAME);
+    }
 }
diff --git a/server/src/main/java/org/opensearch/index/IndexService.java b/server/src/main/java/org/opensearch/index/IndexService.java
@@ -1286,7 +1286,7 @@ AsyncTranslogFSync getFsyncTask() { // for tests
         return fsyncTask;
     }
 
-    AsyncTrimTranslogTask getTrimTranslogTask() { // for tests
+    public AsyncTrimTranslogTask getTrimTranslogTask() { // for tests
         return trimTranslogTask;
     }