[ 
https://issues.apache.org/jira/browse/HADOOP-19862?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18089441#comment-18089441
 ] 

ASF GitHub Bot commented on HADOOP-19862:
-----------------------------------------

ajfabbri commented on code in PR #8426:
URL: https://github.com/apache/hadoop/pull/8426#discussion_r3422798643


##########
hadoop-tools/hadoop-aws/src/test/java/org/apache/hadoop/fs/s3a/TestSharedScheduledExecutor.java:
##########
@@ -0,0 +1,280 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.hadoop.fs.s3a;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.fs.s3a.impl.LazySharedThreadPoolHolder;
+import org.assertj.core.api.Assertions;
+import org.junit.jupiter.api.Test;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.ScheduledExecutorService;
+import java.util.concurrent.ScheduledThreadPoolExecutor;
+
+import static 
org.apache.hadoop.fs.s3a.Constants.AWS_S3_CLIENT_SHARED_THREADPOOL_ENABLED;
+import static 
org.apache.hadoop.fs.s3a.Constants.AWS_S3_CLIENT_SHARED_THREADPOOL_KEEPALIVE;
+import static 
org.apache.hadoop.fs.s3a.Constants.AWS_S3_CLIENT_SHARED_THREADPOOL_SIZE;
+
+/**
+ * Tests for the shared scheduled executors in DefaultS3ClientFactory.
+ */
+public class TestSharedScheduledExecutor {
+
+  @Test
+  public void testLazyHolderDisabledByDefault() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-disabled");
+    Configuration conf = new Configuration();
+    ScheduledExecutorService executor = holder.get(conf);
+    Assertions.assertThat(executor)
+        .as("Executor should be null when disabled")
+        .isNull();
+  }
+
+  @Test
+  public void testCreateScheduledExecutorConfiguration() {
+    ScheduledExecutorService executor =
+        LazySharedThreadPoolHolder.createScheduledExecutor("test-scheduler", 
10, 30);
+    Assertions.assertThat(executor)
+        .as("Executor should be a ScheduledThreadPoolExecutor")
+        .isInstanceOf(ScheduledThreadPoolExecutor.class);
+
+    ScheduledThreadPoolExecutor poolExecutor = (ScheduledThreadPoolExecutor) 
executor;
+    Assertions.assertThat(poolExecutor.getCorePoolSize())
+        .as("Core pool size should be 10")
+        .isEqualTo(10);
+    Assertions.assertThat(poolExecutor.allowsCoreThreadTimeOut())
+        .as("Core threads should be allowed to time out")
+        .isTrue();
+
+    executor.shutdown();
+  }
+
+  @Test
+  public void testCreateScheduledExecutorThreadsAreDaemon() throws Exception {
+    ScheduledExecutorService executor =
+        LazySharedThreadPoolHolder.createScheduledExecutor("test-daemon", 5, 
60);
+    final boolean[] isDaemon = new boolean[1];
+    executor.submit(() -> {
+      isDaemon[0] = Thread.currentThread().isDaemon();
+    }).get();
+    Assertions.assertThat(isDaemon[0])
+        .as("Executor threads should be daemon threads")
+        .isTrue();
+    executor.shutdown();
+  }
+
+  @Test
+  public void testCreateScheduledExecutorThreadName() throws Exception {
+    ScheduledExecutorService executor =
+        LazySharedThreadPoolHolder.createScheduledExecutor("custom-prefix", 5, 
60);
+    final String[] threadName = new String[1];
+    executor.submit(() -> {
+      threadName[0] = Thread.currentThread().getName();
+    }).get();
+    Assertions.assertThat(threadName[0])
+        .as("Thread name should match custom prefix")
+        .startsWith("custom-prefix");
+    executor.shutdown();
+  }
+
+  @Test
+  public void testLazyHolderRejectsNegativePoolSize() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-pool");
+    Configuration conf = new Configuration();
+    conf.setBoolean("test.enabled", true);
+    conf.setInt("test.size", -1);
+    conf.setInt("test.keepalive", 60);
+    Assertions.assertThatThrownBy(() -> holder.get(conf))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("test.size")
+        .hasMessageContaining("must be positive");
+  }
+
+  @Test
+  public void testLazyHolderRejectsZeroPoolSize() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-pool");
+    Configuration conf = new Configuration();
+    conf.setBoolean("test.enabled", true);
+    conf.setInt("test.size", 0);
+    conf.setInt("test.keepalive", 60);
+    Assertions.assertThatThrownBy(() -> holder.get(conf))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("test.size")
+        .hasMessageContaining("must be positive");
+  }
+
+  @Test
+  public void testLazyHolderRejectsNegativeKeepAlive() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-pool");
+    Configuration conf = new Configuration();
+    conf.setBoolean("test.enabled", true);
+    conf.setInt("test.size", 5);
+    conf.setInt("test.keepalive", -1);
+    Assertions.assertThatThrownBy(() -> holder.get(conf))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("test.keepalive")
+        .hasMessageContaining("must be positive");
+  }
+
+  @Test
+  public void testLazyHolderRejectsZeroKeepAlive() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-pool");
+    Configuration conf = new Configuration();
+    conf.setBoolean("test.enabled", true);
+    conf.setInt("test.size", 5);
+    conf.setInt("test.keepalive", 0);
+    Assertions.assertThatThrownBy(() -> holder.get(conf))
+        .isInstanceOf(IllegalArgumentException.class)
+        .hasMessageContaining("test.keepalive")
+        .hasMessageContaining("must be positive");
+  }
+
+  @Test
+  public void testLazyHolderAcceptsValidConfig() {
+    LazySharedThreadPoolHolder holder = new LazySharedThreadPoolHolder(
+        "test.enabled", "test.size", "test.keepalive", "test-valid");
+    Configuration conf = new Configuration();
+    conf.setBoolean("test.enabled", true);
+    conf.setInt("test.size", 5);
+    conf.setInt("test.keepalive", 30);
+    ScheduledExecutorService executor = holder.get(conf);
+    Assertions.assertThat(executor)
+        .as("Executor should be created with valid config")
+        .isNotNull();
+    executor.shutdown();

Review Comment:
   Wouldn't hurt to assert that the values (size, keepalive) are correct 
(sepecially since the constructor takes all strings, so it is easy to misplace 
/ swap arguments). (This comment is a nit: totally optional to address or not)





> S3A: Thread leak from AWS SDK v2 ScheduledExecutorService
> ---------------------------------------------------------
>
>                 Key: HADOOP-19862
>                 URL: https://issues.apache.org/jira/browse/HADOOP-19862
>             Project: Hadoop Common
>          Issue Type: Bug
>            Reporter: Konstantin Bereznyakov
>            Priority: Major
>              Labels: pull-request-available
>
> AWS SDK v2 S3 clients create internal ScheduledExecutorService instances that 
> accumulate over time, causing unbounded thread growth. Thread dumps show 
> thousands of sdk-ScheduledExecutor-* threads in processes that create 
> multiple S3AFileSystem instances.
>   Environment
>   - Hadoop 3.4.x with AWS SDK v2
>   - Not observed in Hadoop 3.3.x (AWS SDK v1)
>   Observed Behavior
>   Thread dump comparison:
>   Hadoop 3.3.x (AWS SDK v1):  Normal thread count
>   Hadoop 3.4.x (AWS SDK v2):  1600+ "sdk-ScheduledExecutor-*" threads
>   Thread pattern:
>   "sdk-ScheduledExecutor-0-0" daemon prio=5 waiting
>   "sdk-ScheduledExecutor-0-1" daemon prio=5 waiting
>   ...
>   "sdk-ScheduledExecutor-0-4" daemon prio=5 waiting
>   "sdk-ScheduledExecutor-1-0" daemon prio=5 waiting
>   ...
>   Root Cause
>   AWS SDK v2's SdkDefaultClientBuilder creates a ScheduledThreadPoolExecutor 
> with 5 threads per client when no executor is explicitly provided 
> (https://github.com/aws/aws-sdk-java-v2/issues/1690):
>   Executors.newScheduledThreadPool(5,
>       new 
> ThreadFactoryBuilder().threadNamePrefix("sdk-ScheduledExecutor").build())
>   These threads are used for retry scheduling, timeout handling, and 
> credential refresh.
>   Contributing Factors
>   1. AbstractFileSystem has no caching
>   Unlike FileSystem.get() which uses CACHE.get(uri, conf), 
> AbstractFileSystem.get() always creates new instances:
>   // AbstractFileSystem.java:263-266
>   public static AbstractFileSystem get(final URI uri, final Configuration 
> conf) {
>       return createFileSystem(uri, conf);  // NO CACHING
>   }
>   Each FileContext.getFileContext() call with an S3 URI creates:
>   - New AbstractFileSystem (S3A)
>   - New S3AFileSystem
>   - New S3Client
>   - 5 new sdk-ScheduledExecutor threads
>   2. S3Client threads not released on close
>   As documented in https://github.com/aws/aws-sdk-java-v2/issues/1690:
>   "When using cached, ephemeral clients, I can see that the scheduled thread 
> pool will at times be leaked when the aws client is evicted"
>   3. Multiple client types affected
>   S3A creates multiple AWS SDK clients:
>   - S3Client (sync)
>   - S3AsyncClient
>   - STS client (for delegation tokens)
>   - KMS client (for encryption)
>   Each client instance creates its own 5-thread pool.
>   Impact
>   - Unbounded thread growth in any process using S3A
>   - Resource exhaustion leading to OOM or system instability
>   - Particularly affects YARN NodeManager, Spark drivers/executors, and other 
> services that create many filesystem instances
>   Related
>   - https://github.com/aws/aws-sdk-java-v2/issues/1690 - SDK issue 
> documenting the problem
>   - https://github.com/aws/aws-sdk-java-v2/pull/4002 - SDK fix allowing 
> shared executor configuration
>   - HADOOP-19624 - Similar thread leak in ABFS (AbfsClientThrottlingAnalyzer)



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to