Skip to content

fix: [Parallel] Fix a flaky parallel test #271

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zhongkechen merged 1 commit into from
Mar 26, 2026
+16 −7
Merged

fix: [Parallel] Fix a flaky parallel test #271

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 16 additions & 7 deletions sdk/src/test/java/software/amazon/lambda/durable/operation/ParallelOperationTest.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@
import software.amazon.awssdk.services.lambda.model.OperationAction;
import software.amazon.awssdk.services.lambda.model.OperationStatus;
import software.amazon.awssdk.services.lambda.model.OperationType;
import software.amazon.awssdk.services.lambda.model.OperationUpdate;
import software.amazon.lambda.durable.DurableConfig;
import software.amazon.lambda.durable.TestUtils;
import software.amazon.lambda.durable.TypeToken;
Expand All @@ -44,14 +45,21 @@ class ParallelOperationTest {

private DurableContextImpl durableContext;
private ExecutionManager executionManager;
// Thread-safe backing store for getOperationAndUpdateReplayState.
// Tests pre-populate this; doAnswer writes here before firing onCheckpointComplete,
// guaranteeing visibility to any thread that reads after the future unblocks.
private ConcurrentHashMap<String, Operation> operationStore;

@BeforeEach
void setUp() {
durableContext = mock(DurableContextImpl.class);
executionManager = mock(ExecutionManager.class);
operationStore = new ConcurrentHashMap<>();

when(executionManager.getCurrentThreadContext()).thenReturn(new ThreadContext(null, ThreadType.CONTEXT));
when(executionManager.getOperationAndUpdateReplayState(anyString())).thenReturn(null);
// Delegate to operationStore so all reads see the latest write, regardless of thread.
when(executionManager.getOperationAndUpdateReplayState(anyString()))
.thenAnswer(inv -> operationStore.get(inv.getArgument(0)));

var childContext = mock(DurableContextImpl.class);
when(childContext.getExecutionManager()).thenReturn(executionManager);
Expand All @@ -77,9 +85,10 @@ void setUp() {
.when(executionManager)
.registerOperation(any());

// Simulate the real backend for all sendOperationUpdate calls:
// - For SUCCEED on the parallel op: update the stub and fire onCheckpointComplete to unblock join().
// - For everything else (START, child checkpoints): just return a completed future.
// Simulate the real backend for all sendOperationUpdate calls.
// For SUCCEED on the parallel op: write to operationStore first (establishes happens-before
// via ConcurrentHashMap's volatile semantics), then fire onCheckpointComplete to unblock join().
// This ordering guarantees getOperationAndUpdateReplayState() never returns null after unblocking.
var succeededParallelOp = Operation.builder()
.id(OPERATION_ID)
.name("test-parallel")
Expand All @@ -88,11 +97,11 @@ void setUp() {
.status(OperationStatus.SUCCEEDED)
.build();
doAnswer(inv -> {
var update = (software.amazon.awssdk.services.lambda.model.OperationUpdate) inv.getArgument(0);
var update = (OperationUpdate) inv.getArgument(0);

if (OPERATION_ID.equals(update.id()) && update.action() == OperationAction.SUCCEED) {
when(executionManager.getOperationAndUpdateReplayState(OPERATION_ID))
.thenReturn(succeededParallelOp);
// Write before completing the future — ConcurrentHashMap guarantees visibility.
operationStore.put(OPERATION_ID, succeededParallelOp);
Copy link
Copy Markdown
Contributor

@zhongkechen zhongkechen Mar 26, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this the only case that is flaky?

Copy link
Copy Markdown
Contributor Author

@wangyb-A wangyb-A Mar 26, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes I believe so, this is the only place we mock the real backend behaviour

var op = registeredOps.get(OPERATION_ID);
if (op != null) {
op.onCheckpointComplete(succeededParallelOp);
Expand Down
Loading