ByteBufferStoringSubscriber.blockingTransferTo hanging indefinitely

[liuchang0520]

Describe the bug

Our application uses SDK v2 s3AsyncClient to send getObject request and chain futures to process the response.
We noticed that the sdk client thread sdk-async-response-* is stuck when reading the InputStream converted by AsyncResponseTransformer.toBlockingInputStream().

Implementation following:

final CompletableFuture<InputStream> future = new CompletableFuture<>();
asyncClient.getObject(requestBuilder.build(), AsyncResponseTransformer.toBlockingInputStream())
                    .handle((responseInputStream, e) -> {
                        if (e != null)
                            // exception handling step
                        else {
                            future.complete(responseInputStream);
                        }
                        return null;
                    });
return future;

Futures to read from inputStream is chained based on this future:

future.thenApply(inputStream -> {
                    ByteBuf buf = ....
                    try {
                        int expectedLen = .....;
                        buf.writeBytes(inputStream, expectedLen);
                        return buf;
                    } catch (Exception e) {
                       ......
                    } finally {
                        ......
                    }
                });

Stack trace following:

"sdk-async-response-1-883" #17464 [313453] daemon prio=5 os_prio=0 cpu=8547.18ms elapsed=168400.23s tid=0x0000ffff4c039ba0 nid=313453 waiting on condition  [0x0000fff939a53000]
   java.lang.Thread.State: WAITING (parking)
	at jdk.internal.misc.Unsafe.park(java.base@21.0.4/Native Method)
	- parking to wait for  <0x000000071ad09a68> (a java.util.concurrent.Phaser$QNode)
	at java.util.concurrent.locks.LockSupport.park(java.base@21.0.4/LockSupport.java:221)
	at java.util.concurrent.Phaser$QNode.block(java.base@21.0.4/Phaser.java:1133)
	at java.util.concurrent.ForkJoinPool.unmanagedBlock(java.base@21.0.4/ForkJoinPool.java:3780)
	at java.util.concurrent.ForkJoinPool.managedBlock(java.base@21.0.4/ForkJoinPool.java:3725)
	at java.util.concurrent.Phaser.internalAwaitAdvance(java.base@21.0.4/Phaser.java:1063)
	at java.util.concurrent.Phaser.awaitAdvanceInterruptibly(java.base@21.0.4/Phaser.java:753)
	at software.amazon.awssdk.utils.async.ByteBufferStoringSubscriber.blockingTransferTo(ByteBufferStoringSubscriber.java:148)
	at software.amazon.awssdk.utils.async.InputStreamSubscriber.read(InputStreamSubscriber.java:134)
	at software.amazon.awssdk.http.async.AbortableInputStreamSubscriber.read(AbortableInputStreamSubscriber.java:67)
	at software.amazon.awssdk.core.io.SdkFilterInputStream.read(SdkFilterInputStream.java:66)
       ....
	at io.netty.buffer.UnsafeByteBufUtil.setBytes(UnsafeByteBufUtil.java:472)
	at io.netty.buffer.PooledUnsafeDirectByteBuf.setBytes(PooledUnsafeDirectByteBuf.java:211)
	at io.netty.buffer.AbstractByteBuf.writeBytes(AbstractByteBuf.java:1122)

        **....... proprietary code above to invoke asyncClient.getObject and chain future to process the response ........**

	at java.util.concurrent.CompletableFuture.uniHandle(java.base@21.0.4/CompletableFuture.java:934)
	at java.util.concurrent.CompletableFuture$UniHandle.tryFire(java.base@21.0.4/CompletableFuture.java:911)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallMetricCollectionStage.lambda$execute$0(AsyncApiCallMetricCollectionStage.java:58)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallMetricCollectionStage$$Lambda/0x00000018027ad898.accept(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniWhenComplete(java.base@21.0.4/CompletableFuture.java:863)
	at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(java.base@21.0.4/CompletableFuture.java:841)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallTimeoutTrackingStage.lambda$execute$2(AsyncApiCallTimeoutTrackingStage.java:69)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallTimeoutTrackingStage$$Lambda/0x00000018027ad668.accept(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniWhenComplete(java.base@21.0.4/CompletableFuture.java:863)
	at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(java.base@21.0.4/CompletableFuture.java:841)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncRetryableStage2$RetryingExecutor.lambda$attemptExecute$1(AsyncRetryableStage2.java:128)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncRetryableStage2$RetryingExecutor$$Lambda/0x00000018027aef08.accept(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniWhenComplete(java.base@21.0.4/CompletableFuture.java:863)
	at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(java.base@21.0.4/CompletableFuture.java:841)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.MakeAsyncHttpRequestStage.lambda$execute$0(MakeAsyncHttpRequestStage.java:110)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.MakeAsyncHttpRequestStage$$Lambda/0x00000018027ae8a8.accept(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniWhenComplete(java.base@21.0.4/CompletableFuture.java:863)
	at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(java.base@21.0.4/CompletableFuture.java:841)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.MakeAsyncHttpRequestStage.completeResponseFuture(MakeAsyncHttpRequestStage.java:253)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.MakeAsyncHttpRequestStage.lambda$executeHttpRequest$3(MakeAsyncHttpRequestStage.java:167)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.MakeAsyncHttpRequestStage$$Lambda/0x00000018027ae448.apply(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniHandle(java.base@21.0.4/CompletableFuture.java:934)
	at java.util.concurrent.CompletableFuture$UniHandle.tryFire(java.base@21.0.4/CompletableFuture.java:911)
	at java.util.concurrent.CompletableFuture$Completion.run(java.base@21.0.4/CompletableFuture.java:482)
	at java.util.concurrent.ThreadPoolExecutor.runWorker(java.base@21.0.4/ThreadPoolExecutor.java:1144)
	at java.util.concurrent.ThreadPoolExecutor$Worker.run(java.base@21.0.4/ThreadPoolExecutor.java:642)
	at java.lang.Thread.runWith(java.base@21.0.4/Thread.java:1596)
	at java.lang.Thread.run(java.base@21.0.4/Thread.java:1583)

A couple of questions:

1. What could cause the indefinite blocking when reading from the inputStream ?
2. How can we avoid the indefinite blocking when processing the inputStream response from s3AsyncClient.getObject:
   e.g. Is there a way to specify timeout when using the S3AsyncClient to avoid indefinite blocking ?
   Could we expose exception so we can handle it from client side to avoid the indefinite blocking ?
3. In the stack trace, we also noticed some timeout and retry - would could be the cause of the timeout:

software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallTimeoutTrackingStage.lambda$execute$2(AsyncApiCallTimeoutTrackingStage.java:69)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncApiCallTimeoutTrackingStage$$Lambda/0x00000018027ad668.accept(Unknown Source)
	at java.util.concurrent.CompletableFuture.uniWhenComplete(java.base@21.0.4/CompletableFuture.java:863)
	at java.util.concurrent.CompletableFuture$UniWhenComplete.tryFire(java.base@21.0.4/CompletableFuture.java:841)
	at java.util.concurrent.CompletableFuture.postComplete(java.base@21.0.4/CompletableFuture.java:510)
	at java.util.concurrent.CompletableFuture.complete(java.base@21.0.4/CompletableFuture.java:2179)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncRetryableStage2$RetryingExecutor.lambda$attemptExecute$1(AsyncRetryableStage2.java:128)
	at software.amazon.awssdk.core.internal.http.pipeline.stages.AsyncRetryableStage2$RetryingExecutor$$Lambda/0x00000018027aef08.accept(Unknown Source)

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

We expect the reading from inputStream to be successful into the destination buffer.

Current Behavior

The inputStream reading blocks indefinitely(software.amazon.awssdk.utils.async.ByteBufferStoringSubscriber.blockingTransferTo)

Reproduction Steps

Same as the code snippet above.

Possible Solution

No response

Additional Information/Context

No response

AWS Java SDK version used

2.27.9

JDK version used

Java 21

Operating System and version

Linux Wolfi, aarch64

[liuchang0520]

Note that AsyncResponseTransformer.toBytes() doesn't apply to our use case due to a known memory issue in SDK v2:

1. https://github.com/aws/aws-sdk-java-v2/issues/4392
2. S3 getObject combined with AsyncResponseTransformer.toBytes() copies too much data #3193

[debora-ito]

Hi @liuchang0520, the support case you had with us got resolved, so I'm marking this issue to auto close soon.

One recommendation is to make sure to close the inputstream after the stream reading is finished.

If you (or anyone else) are still experiencing this same hanging issue, please let us know, ideally also providing a self-contained repro code (we're not able to reproduce the hanging in our side).

[liuchang0520]

The inputStream is correctly closed on our application side.
Any update regarding why the read is blocked indefinitely?

[liuchang0520]

Please disregard the support case as it doesn't really capture the issue.