Merge remote-tracking branch 'origin/main' into rupei/kv-indexer-python-launcher-pr

Signed-off-by: PeaBrane <yanrpei@gmail.com>

# Conflicts:
#	lib/bindings/python/Cargo.toml
#	lib/bindings/python/rust/lib.rs
#	tests/router/test_router_e2e_with_mockers.py

Author: PeaBrane

README.md

@@ -50,7 +50,6 @@ Built in Rust for performance and Python for extensibility, Dynamo is fully open
 
 | | [SGLang](docs/backends/sglang/README.md) | [TensorRT-LLM](docs/backends/trtllm/README.md) | [vLLM](docs/backends/vllm/README.md) |
 |---|:----:|:----------:|:--:|
-| **Best For** | High-throughput serving | Maximum performance | Broadest feature coverage |
 | [**Disaggregated Serving**](docs/design-docs/disagg-serving.md) | ✅ | ✅ | ✅ |
 | [**KV-Aware Routing**](docs/components/router/README.md) | ✅ | ✅ | ✅ |
 | [**SLA-Based Planner**](docs/components/planner/planner-guide.md) | ✅ | ✅ | ✅ |

container/templates/wheel_builder.Dockerfile

@@ -439,9 +439,9 @@ RUN --mount=type=secret,id=aws-key-id,env=AWS_ACCESS_KEY_ID \
     uv build --wheel --out-dir /opt/dynamo/dist && \
     cd /opt/dynamo/lib/bindings/python && \
     if [ "$ENABLE_MEDIA_FFMPEG" = "true" ]; then \
-        maturin build --release --features "media-ffmpeg,kv-indexer" --out /opt/dynamo/dist; \
+        maturin build --release --features "media-ffmpeg" --out /opt/dynamo/dist; \
     else \
-        maturin build --release --features "kv-indexer" --out /opt/dynamo/dist; \
+        maturin build --release --out /opt/dynamo/dist; \
     fi && \
     /tmp/use-sccache.sh show-stats "Dynamo Runtime"
 

lib/bindings/python/Cargo.lock

@@ -31,6 +31,9 @@ mod address;
 
 pub mod tcp;
 
+#[cfg(unix)]
+pub mod uds;
+
 // #[cfg(feature = "ucx")]
 // pub mod ucx;
 

lib/velo-transports/src/lib.rs

@@ -392,9 +392,9 @@ async fn connection_writer_task(
     instance_id: crate::InstanceId,
     rx: flume::Receiver<SendTask>,
     connections: Arc<DashMap<crate::InstanceId, ConnectionHandle>>,
-    _cancel_token: CancellationToken,
+    cancel_token: CancellationToken,
 ) -> Result<()> {
-    let result = connection_writer_inner(addr, instance_id, &rx).await;
+    let result = connection_writer_inner(addr, instance_id, &rx, &cancel_token).await;
 
     // Always drain queued messages and notify their error handlers.
     //
@@ -426,10 +426,14 @@ async fn connection_writer_inner(
     addr: SocketAddr,
     instance_id: crate::InstanceId,
     rx: &flume::Receiver<SendTask>,
+    cancel_token: &CancellationToken,
 ) -> Result<()> {
     debug!("Connecting to {}", addr);
 
-    let mut stream = TcpStream::connect(addr).await.context("connect failed")?;
+    let mut stream = tokio::select! {
+        _ = cancel_token.cancelled() => return Ok(()),
+        res = TcpStream::connect(addr) => res.context("connect failed")?,
+    };
 
     if let Err(e) = stream.set_nodelay(true) {
         warn!("Failed to set TCP_NODELAY: {}", e);
@@ -450,7 +454,14 @@ async fn connection_writer_inner(
 
     debug!("Connected to {}", addr);
 
-    while let Ok(msg) = rx.recv_async().await {
+    loop {
+        let msg = tokio::select! {
+            _ = cancel_token.cancelled() => break,
+            res = rx.recv_async() => match res {
+                Ok(msg) => msg,
+                Err(_) => break,
+            },
+        };
         if let Err(e) =
             TcpFrameCodec::encode_frame(&mut stream, msg.msg_type, &msg.header, &msg.payload).await
         {