From a5d8fa3ae1296867caf78dc060eb27902b651fca Mon Sep 17 00:00:00 2001
From: Derek Cofausper <256792747+decofe@users.noreply.github.com>
Date: Tue, 3 Mar 2026 06:44:43 -0800
Subject: [PATCH] feat(metrics): add /debug/tokio/dump endpoint for tokio task
 dumps (#22737)

Co-authored-by: Alexey Shekhirin <5773434+shekhirin@users.noreply.github.com>
---
 Cargo.toml                        |  1 +
 crates/node/metrics/src/server.rs | 37 ++++++++++++++++++++++++++++---
 2 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/Cargo.toml b/Cargo.toml
index 79e93eb6a9..12ceb56b73 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -170,6 +170,7 @@ rust.rust_2018_idioms = { level = "deny", priority = -1 }
 rust.unreachable_pub = "warn"
 rust.unused_must_use = "deny"
 rust.rust_2024_incompatible_pat = "warn"
+rust.unexpected_cfgs = { level = "warn", check-cfg = ['cfg(tokio_unstable)'] }
 rustdoc.all = "warn"
 # rust.unnameable-types = "warn"
 
diff --git a/crates/node/metrics/src/server.rs b/crates/node/metrics/src/server.rs
index 23561c575f..56927ece36 100644
--- a/crates/node/metrics/src/server.rs
+++ b/crates/node/metrics/src/server.rs
@@ -148,8 +148,13 @@ impl MetricServer {
             let hook = hook.clone();
             let pprof_dump_dir = pprof_dump_dir.clone();
             let service = tower::service_fn(move |req: Request<_>| {
-                let response = handle_request(req.uri().path(), &*hook, handle, &pprof_dump_dir);
-                async move { Ok::<_, Infallible>(response) }
+                let hook = hook.clone();
+                let pprof_dump_dir = pprof_dump_dir.clone();
+                async move {
+                    let response =
+                        handle_request(req.uri().path(), &*hook, handle, &pprof_dump_dir).await;
+                    Ok::<_, Infallible>(response)
+                }
             });
 
             let mut shutdown = signal.clone().ignore_guard();
@@ -307,7 +312,7 @@ fn describe_io_stats() {
 #[cfg(not(target_os = "linux"))]
 const fn describe_io_stats() {}
 
-fn handle_request(
+async fn handle_request(
     path: &str,
     hook: impl Fn(),
     handle: &crate::recorder::PrometheusRecorder,
@@ -315,6 +320,7 @@ fn handle_request(
 ) -> Response<Full<Bytes>> {
     match path {
         "/debug/pprof/heap" => handle_pprof_heap(pprof_dump_dir),
+        "/debug/tokio/dump" => handle_tokio_dump().await,
         _ => {
             hook();
             let metrics = handle.handle().render();
@@ -404,6 +410,31 @@ fn handle_pprof_heap(_pprof_dump_dir: &PathBuf) -> Response<Full<Bytes>> {
     response
 }
 
+#[cfg(tokio_unstable)]
+async fn handle_tokio_dump() -> Response<Full<Bytes>> {
+    let handle = tokio::runtime::Handle::current();
+    let dump = handle.dump().await;
+
+    let mut output = String::new();
+    for (i, task) in dump.tasks().iter().enumerate() {
+        let trace = task.trace();
+        output.push_str(&format!("task {i}:\n{trace}\n\n"));
+    }
+
+    let mut response = Response::new(Full::new(Bytes::from(output)));
+    response.headers_mut().insert(CONTENT_TYPE, HeaderValue::from_static("text/plain"));
+    response
+}
+
+#[cfg(not(tokio_unstable))]
+async fn handle_tokio_dump() -> Response<Full<Bytes>> {
+    let mut response = Response::new(Full::new(Bytes::from_static(
+        b"tokio task dump not available. Rebuild with RUSTFLAGS=\"--cfg tokio_unstable\" and tokio's `taskdump` feature.",
+    )));
+    *response.status_mut() = StatusCode::NOT_IMPLEMENTED;
+    response
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;