Skip to content

[bug]:TaskRuntime.stateLock and TableOptimizingProcess.lock can cause deadlocks. #4002

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

Open
wardlican wants to merge 1 commit into
base: master
Choose a base branch
from
Open

[bug]:TaskRuntime.stateLock and TableOptimizingProcess.lock can cause deadlocks. #4002

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
24 changes: 21 additions & 3 deletions amoro-ams/src/main/java/org/apache/amoro/server/optimizing/OptimizingQueue.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -564,9 +564,16 @@ private void ackTask(OptimizingTaskId taskId, OptimizerThread thread) {

private void completeTask(OptimizerThread thread, OptimizingTaskResult result) {
TaskRuntime<?> taskRuntime = getTaskRuntime(result.getTaskId());
// Complete taskRuntime first (this acquires the database lock, but not under
// TableOptimizingProcess.this.lock protection)
// This avoids deadlock by ensuring consistent lock acquisition order (acquire
// TableOptimizingProcess.this.lock first, then database lock)
taskRuntime.complete(thread, result);
// After taskRuntime.complete() returns (database lock is released), acquire
// TableOptimizingProcess.this.lock and call acceptResult
lock.lock();
try {
taskRuntime.complete(thread, result);
acceptResult(taskRuntime);
} finally {
lock.unlock();
}
Expand Down Expand Up @@ -891,7 +898,14 @@ private void persistAndSetCompleted(boolean success) {
}

private void cancelTasks() {
// Cancel all tasks first (this acquires the database lock, but not under
// TableOptimizingProcess.this.lock protection)
taskMap.values().forEach(TaskRuntime::tryCanceling);
// Then call acceptResult for each canceled task (while holding
// TableOptimizingProcess.this.lock)
// This avoids deadlock by ensuring consistent lock acquisition order (acquire
// TableOptimizingProcess.this.lock first, then database lock)
taskMap.values().forEach(this::acceptResult);
}

private void loadTaskRuntimes(OptimizingProcess optimizingProcess) {
Expand All @@ -905,7 +919,9 @@ private void loadTaskRuntimes(OptimizingProcess optimizingProcess) {
Map<Integer, RewriteFilesInput> inputs = TaskFilesPersistence.loadTaskInputs(processId);
taskRuntimes.forEach(
taskRuntime -> {
taskRuntime.getCompletedFuture().whenCompleted(() -> acceptResult(taskRuntime));
// Remove whenCompleted callback registration to avoid deadlock caused by
// synchronously calling acceptResult while holding database lock
// acceptResult will be explicitly called in completeTask
taskRuntime
.getTaskDescriptor()
.setInput(inputs.get(taskRuntime.getTaskId().getTaskId()));
Expand Down Expand Up @@ -935,7 +951,9 @@ private void loadTaskRuntimes(List<RewriteStageTask> taskDescriptors) {
tableRuntime.getTableIdentifier(),
taskRuntime.getTaskId(),
taskRuntime.getSummary());
taskRuntime.getCompletedFuture().whenCompleted(() -> acceptResult(taskRuntime));
// Remove whenCompleted callback registration to avoid deadlock caused by synchronously
// calling acceptResult while holding database lock
// acceptResult will be explicitly called in completeTask
taskMap.put(taskRuntime.getTaskId(), taskRuntime);
taskQueue.offer(taskRuntime);
}
Expand Down