[Improvement] Avoid potential OOM in SessionPoolExample by replacing unbounded thread pool #17016
Description
Search before asking
- I searched in the issues and found nothing similar.
Version
Master
Describe the bug and provide the minimal reproduce step
Description:
In example/session/src/main/java/org/apache/iotdb/SessionPoolExample.java, the ExecutorService is initialized using Executors.newFixedThreadPool(10) (Line 74).
Problem:
Executors.newFixedThreadPool uses an unbounded LinkedBlockingQueue (capacity: Integer.MAX_VALUE) by default. As this is an official example, users often copy-paste this code for production. In high-throughput write scenarios, if the production rate exceeds the consumption rate, tasks accumulate indefinitely in the queue, leading to OutOfMemoryError: Java heap space.
Code Location:
// org.apache.iotdb.SessionPoolExample.java
service = Executors.newFixedThreadPool(10); // Unbounded Queue risk
What did you expect to see?
Official examples should demonstrate best practices by using bounded queues to ensure system stability. The thread pool should provide backpressure (block or reject) when the queue is full, preventing memory exhaustion.
What did you see instead?
The example uses an unbounded queue pattern (UBSCQ), which creates a hidden memory leak risk for downstream users who adopt this code snippet.
Anything else?
Suggested Fix: Replace the factory method with a custom ThreadPoolExecutor using a bounded queue (e.g., ArrayBlockingQueue).
Proposed Code:
service = new ThreadPoolExecutor(
10, 10, 0L, TimeUnit.MILLISECONDS,
new ArrayBlockingQueue<>(1000) // Bounded capacity
);
Are you willing to submit a PR?
- I'm willing to submit a PR!