Fix flaky behavior of RemoteStoreRestoreIT.testRateLimitedRemoteDownl…

…oads (opensearch-project#9887) * Fix flaky behavior of RemoteStoreRestoreIT.testRateLimitedRemoteDownloads Signed-off-by: Sachin Kale <kalsac@amazon.com> Signed-off-by: Shivansh Arora <hishiv@amazon.com>
Gaurav614 · Apr 25, 2024 · b51dde9 · b51dde9
1 parent 8a52988
commit b51dde9
Show file tree

Hide file tree

Showing 2 changed files with 4 additions and 3 deletions.
diff --git a/...src/internalClusterTest/java/org/opensearch/remotestore/RemoteStoreBaseIntegTestCase.java b/...src/internalClusterTest/java/org/opensearch/remotestore/RemoteStoreBaseIntegTestCase.java
@@ -230,7 +230,7 @@ public static Settings buildRemoteStoreNodeAttributes(
 
         if (withRateLimiterAttributes) {
             settings.put(segmentRepoSettingsAttributeKeyPrefix + "compress", randomBoolean())
-                .put(segmentRepoSettingsAttributeKeyPrefix + "max_remote_download_bytes_per_sec", "2kb")
+                .put(segmentRepoSettingsAttributeKeyPrefix + "max_remote_download_bytes_per_sec", "4kb")
                 .put(segmentRepoSettingsAttributeKeyPrefix + "chunk_size", 200, ByteSizeUnit.BYTES);
         }
 

diff --git a/server/src/internalClusterTest/java/org/opensearch/remotestore/RemoteStoreRestoreIT.java b/server/src/internalClusterTest/java/org/opensearch/remotestore/RemoteStoreRestoreIT.java
@@ -406,9 +406,10 @@ public void testRateLimitedRemoteDownloads() throws Exception {
             for (RepositoriesService repositoriesService : internalCluster().getDataNodeInstances(RepositoriesService.class)) {
                 downloadPauseTime += repositoriesService.repository(REPOSITORY_NAME).getRemoteDownloadThrottleTimeInNanos();
             }
-            assertThat(downloadPauseTime, greaterThan(TimeValue.timeValueSeconds(randomIntBetween(5, 10)).nanos()));
+            assertThat(downloadPauseTime, greaterThan(TimeValue.timeValueSeconds(randomIntBetween(3, 5)).nanos()));
         }, 30, TimeUnit.SECONDS);
-        ensureGreen(INDEX_NAME);
+        // Waiting for extended period for green state so that rate limit does not cause flakiness
+        ensureGreen(TimeValue.timeValueSeconds(120), INDEX_NAME);
         // This is required to get updated number from already active shards which were not restored
         assertEquals(shardCount, getNumShards(INDEX_NAME).totalNumShards);
         assertEquals(0, getNumShards(INDEX_NAME).numReplicas);