On problems with threads in node.js

The article argues that Node.js is single-threaded only at the JavaScript level, while asynchronous work is often handled by libuv’s thread pool. A small example with delayed directory reads shows that once the default pool of four threads is saturated, later requests wait in the queue and finish later.

Reading notes

#

• JavaScript code runs on one thread, but Node.js uses threads for asynchronous work under the hood.
• The example delays scandir() with a shared library so the queueing effect becomes visible.
• libuv submits work requests with a function to run, data for that function, and a function that collects the result back on the main thread.
• The default libuv thread pool size is 4, which explains why six readdir() calls split into two waves.
• The thread pool is used mainly for file system operations and user code.
• In a database-heavy application, query execution can also consume the same pool, so slow queries can block unrelated asynchronous tasks.
• UV_THREADPOOL_SIZE can change the pool size, but only before the first work request creates the pool.
• The article ends with a bug where query termination requests failed because they were queued behind long-running queries.