feature(core): default tokio worker threads to std::thread::available_parallelism() (vectordotdev#18541)

* default tokio worker threads to std::thread::available_parallelism()

* update log

* rename log field

* add upgrade guide entry

* fmt

Author: dsmith3197

src/app.rs

@@ -467,21 +467,17 @@ pub fn build_runtime(threads: Option<usize>, thread_name: &str) -> Result<Runtim
     rt_builder.max_blocking_threads(20_000);
     rt_builder.enable_all().thread_name(thread_name);
 
-    if let Some(threads) = threads {
-        if threads < 1 {
-            #[allow(clippy::print_stderr)]
-            {
-                eprintln!("The `threads` argument must be greater or equal to 1.");
-            }
-            return Err(exitcode::CONFIG);
-        } else {
-            WORKER_THREADS
-                .set(NonZeroUsize::new(threads).expect("already checked"))
-                .expect("double thread initialization");
-            rt_builder.worker_threads(threads);
-        }
+    let threads = threads.unwrap_or_else(crate::num_threads);
+    if threads < 1 {
+        error!("The `threads` argument must be greater or equal to 1.");
+        return Err(exitcode::CONFIG);
     }
+    WORKER_THREADS
+        .set(NonZeroUsize::new(threads).expect("already checked"))
+        .expect("double thread initialization");
+    rt_builder.worker_threads(threads);
 
+    debug!(messaged = "Building runtime.", worker_threads = threads);
     Ok(rt_builder.build().expect("Unable to create async runtime"))
 }
 

website/content/en/highlights/2023-09-26-0-33-0-upgrade-guide.md

@@ -24,3 +24,16 @@ We cover them below to help you upgrade quickly:
 The `armv7` rpm package, `vector-<version>-1.armv7.rpm`, is now published as
 `vector-<version>-1.armv7hl.rpm` to better follow rpm guidelines. The `armv7`
 package will be no longer be published beginning in the 0.34.0 release.
+
+### Potentially impactful changes
+
+#### Async runtime default number of worker threads {#runtime-worker-threads}
+
+We've changed the default number of worker threads spawned by Vector's async runtime
+from being the number of CPUs on the host machine to the value returned by
+[`std::thread::available_parallelism()`](https://doc.rust-lang.org/stable/std/thread/fn.available_parallelism.html).
+This should be a better default value for containerized environments where the container
+has limited quotas, but note this change may impact performance.
+
+The number of worker threads used can be seen by enabling debug logging, and the value can
+be overriden by setting the `VECTOR_THREADS` environment variable.