client.rs 33.8 KB
Newer Older
1
// SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2
// SPDX-License-Identifier: Apache-2.0
Ryan Olson's avatar
Ryan Olson committed
3
4
5
6

use std::sync::Arc;

use futures::{SinkExt, StreamExt};
7
8
9
10
11
12
use tokio::io::{AsyncReadExt, ReadHalf, WriteHalf};
use tokio::{
    io::AsyncWriteExt,
    net::TcpStream,
    time::{self, Duration, Instant},
};
Ryan Olson's avatar
Ryan Olson committed
13
14
use tokio_util::codec::{FramedRead, FramedWrite};

15
16
use prometheus::IntCounter;

Ryan Olson's avatar
Ryan Olson committed
17
18
19
use super::{CallHomeHandshake, ControlMessage, TcpStreamConnectionInfo};
use crate::engine::AsyncEngineContext;
use crate::pipeline::network::{
20
    ConnectionInfo, ResponseStreamPrologue, StreamSender,
Ryan Olson's avatar
Ryan Olson committed
21
22
    codec::{TwoPartCodec, TwoPartMessage},
    tcp::StreamType,
23
};
24
use anyhow::{Context, Result, anyhow as error}; // Import SinkExt to use the `send` method
Ryan Olson's avatar
Ryan Olson committed
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

#[allow(dead_code)]
pub struct TcpClient {
    worker_id: String,
}

impl Default for TcpClient {
    fn default() -> Self {
        TcpClient {
            worker_id: uuid::Uuid::new_v4().to_string(),
        }
    }
}

impl TcpClient {
    pub fn new(worker_id: String) -> Self {
        TcpClient { worker_id }
    }

44
    async fn connect(address: &str) -> std::io::Result<TcpStream> {
45
        // try to connect to the address; retry with linear backoff if AddrNotAvailable
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
        let backoff = std::time::Duration::from_millis(200);
        loop {
            match TcpStream::connect(address).await {
                Ok(socket) => {
                    socket.set_nodelay(true)?;
                    return Ok(socket);
                }
                Err(e) => {
                    if e.kind() == std::io::ErrorKind::AddrNotAvailable {
                        tracing::warn!("retry warning: failed to connect: {:?}", e);
                        tokio::time::sleep(backoff).await;
                    } else {
                        return Err(e);
                    }
                }
            }
        }
Ryan Olson's avatar
Ryan Olson committed
63
64
    }

65
    pub async fn create_response_stream(
Ryan Olson's avatar
Ryan Olson committed
66
67
        context: Arc<dyn AsyncEngineContext>,
        info: ConnectionInfo,
68
        cancellation_counter: Option<IntCounter>,
69
70
71
    ) -> Result<StreamSender> {
        let info =
            TcpStreamConnectionInfo::try_from(info).context("tcp-stream-connection-info-error")?;
Ryan Olson's avatar
Ryan Olson committed
72
73
74
        tracing::trace!("Creating response stream for {:?}", info);

        if info.stream_type != StreamType::Response {
75
            return Err(error!(
Ryan Olson's avatar
Ryan Olson committed
76
77
78
79
80
81
                "Invalid stream type; TcpClient requires the stream type to be `response`; however {:?} was passed",
                info.stream_type
            ));
        }

        if info.context != context.id() {
82
            return Err(error!(
Ryan Olson's avatar
Ryan Olson committed
83
84
85
86
87
88
89
                "Invalid context; TcpClient requires the context to be {:?}; however {:?} was passed",
                context.id(),
                info.context
            ));
        }

        let stream = TcpClient::connect(&info.address).await?;
90
        let peer_port = stream.peer_addr().ok().map(|addr| addr.port());
Ryan Olson's avatar
Ryan Olson committed
91
92
        let (read_half, write_half) = tokio::io::split(stream);

93
        let framed_reader = FramedRead::new(read_half, TwoPartCodec::default());
Ryan Olson's avatar
Ryan Olson committed
94
95
96
97
98
99
100
        let mut framed_writer = FramedWrite::new(write_half, TwoPartCodec::default());

        // this is a oneshot channel that will be used to signal when the stream is closed
        // when the stream sender is dropped, the bytes_rx will be closed and the forwarder task will exit
        // the forwarder task will capture the alive_rx half of the oneshot channel; this will close the alive channel
        // so the holder of the alive_tx half will be notified that the stream is closed; the alive_tx channel will be
        // captured by the monitor task
101
102
        let (alive_tx, alive_rx) = tokio::sync::oneshot::channel::<()>();

103
104
105
106
107
108
        let reader_task = tokio::spawn(handle_reader(
            framed_reader,
            context.clone(),
            alive_tx,
            cancellation_counter,
        ));
Ryan Olson's avatar
Ryan Olson committed
109
110
111

        // transport specific handshake message
        let handshake = CallHomeHandshake {
112
            subject: info.subject.clone(),
Ryan Olson's avatar
Ryan Olson committed
113
114
115
            stream_type: StreamType::Response,
        };

116
117
118
        let handshake_bytes = match serde_json::to_vec(&handshake) {
            Ok(hb) => hb,
            Err(err) => {
119
                return Err(error!(
120
                    "create_response_stream: Error converting CallHomeHandshake to JSON array: {err:#}"
121
122
123
                ));
            }
        };
Ryan Olson's avatar
Ryan Olson committed
124
125
126
127
128
129
        let msg = TwoPartMessage::from_header(handshake_bytes.into());

        // issue the the first tcp handshake message
        framed_writer
            .send(msg)
            .await
130
            .map_err(|e| error!("failed to send handshake: {:?}", e))?;
Ryan Olson's avatar
Ryan Olson committed
131
132

        // set up the channel to send bytes to the transport layer
133
        let (bytes_tx, bytes_rx) = tokio::sync::mpsc::channel(64);
Ryan Olson's avatar
Ryan Olson committed
134
135

        // forwards the bytes send from this stream to the transport layer; hold the alive_rx half of the oneshot channel
136

137
        let writer_task = tokio::spawn(handle_writer(framed_writer, bytes_rx, alive_rx, context));
138

139
        let subject = info.subject.clone();
140
141
142
143
144
145
146
147
        tokio::spawn(async move {
            // await both tasks
            let (reader, writer) = tokio::join!(reader_task, writer_task);

            match (reader, writer) {
                (Ok(reader), Ok(writer)) => {
                    let reader = reader.into_inner();

148
149
150
151
152
153
154
155
                    let writer = match writer {
                        Ok(writer) => writer.into_inner(),
                        Err(e) => {
                            tracing::error!("failed to join writer task: {:?}", e);
                            return Err(e);
                        }
                    };

156
157
                    let mut stream = reader.unsplit(writer);

158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
                    // await the tcp server to shutdown the socket connection
                    // set a timeout for the server shutdown
                    let mut buf = vec![0u8; 1024];
                    let deadline = Instant::now() + Duration::from_secs(10);
                    loop {
                        let n = time::timeout_at(deadline, stream.read(&mut buf))
                            .await
                            .inspect_err(|_| {
                                tracing::debug!("server did not close socket within the deadline");
                            })?
                            .inspect_err(|e| {
                                tracing::debug!("failed to read from stream: {:?}", e);
                            })?;
                        if n == 0 {
                            // Server has closed (FIN)
                            break;
                        }
                    }

                    Ok(())
178
                }
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
                (Err(reader_err), Ok(_)) => {
                    tracing::error!(
                        "reader task failed to join (peer_port: {peer_port:?}, subject: {subject}): {reader_err:?}"
                    );
                    anyhow::bail!(
                        "reader task failed to join (peer_port: {peer_port:?}, subject: {subject}): {reader_err:?}"
                    );
                }
                (Ok(_), Err(writer_err)) => {
                    tracing::error!(
                        "writer task failed to join (peer_port: {peer_port:?}, subject: {subject}): {writer_err:?}"
                    );
                    anyhow::bail!(
                        "writer task failed to join (peer_port: {peer_port:?}, subject: {subject}): {writer_err:?}"
                    );
                }
                (Err(reader_err), Err(writer_err)) => {
                    tracing::error!(
                        "both reader and writer tasks failed to join (peer_port: {peer_port:?}, subject: {subject}) - reader: {reader_err:?}, writer: {writer_err:?}"
                    );
                    anyhow::bail!(
                        "both reader and writer tasks failed to join (peer_port: {peer_port:?}, subject: {subject}) - reader: {reader_err:?}, writer: {writer_err:?}"
                    );
202
203
204
                }
            }
        });
Ryan Olson's avatar
Ryan Olson committed
205
206
207
208
209
210
211
212
213
214
215
216
217
218

        // set up the prologue for the stream
        // this might have transport specific metadata in the future
        let prologue = Some(ResponseStreamPrologue { error: None });

        // create the stream sender
        let stream_sender = StreamSender {
            tx: bytes_tx,
            prologue,
        };

        Ok(stream_sender)
    }
}
219

220
221
async fn handle_reader(
    framed_reader: FramedRead<tokio::io::ReadHalf<tokio::net::TcpStream>, TwoPartCodec>,
222
    context: Arc<dyn AsyncEngineContext>,
223
    alive_tx: tokio::sync::oneshot::Sender<()>,
224
    cancellation_counter: Option<IntCounter>,
225
226
227
) -> FramedRead<tokio::io::ReadHalf<tokio::net::TcpStream>, TwoPartCodec> {
    let mut framed_reader = framed_reader;
    let mut alive_tx = alive_tx;
228
    let mut cancellation_counted = false;
229
230
231
232
233
234
235
    loop {
        tokio::select! {
            msg = framed_reader.next() => {
                match msg {
                    Some(Ok(two_part_msg)) => {
                        match two_part_msg.optional_parts() {
                           (Some(bytes), None) => {
236
                                let msg = match serde_json::from_slice::<ControlMessage>(bytes) {
237
                                    Ok(msg) => msg,
238
239
                                    Err(_) => {
                                        // TODO(#171) - address fatal errors
240
                                        panic!("fatal error - invalid control message detected");
241
242
                                    }
                                };
243

244
245
                                match msg {
                                    ControlMessage::Stop => {
246
247
248
249
                                        if let Some(counter) = &cancellation_counter && !cancellation_counted {
                                            counter.inc();
                                            cancellation_counted = true;
                                        }
250
251
252
                                        context.stop();
                                    }
                                    ControlMessage::Kill => {
253
254
255
256
                                        if let Some(counter) = &cancellation_counter && !cancellation_counted {
                                            counter.inc();
                                            cancellation_counted = true;
                                        }
257
                                        context.kill();
258
259
260
261
                                    }
                                    ControlMessage::Sentinel => {
                                        // TODO(#171) - address fatal errors
                                        panic!("received a sentinel message; this should never happen");
262
263
264
265
                                    }
                                }
                           }
                           _ => {
266
                                panic!("received a non-control message; this should never happen");
267
268
269
                           }
                        }
                    }
270
                    Some(Err(e)) => {
271
272
                        // TODO(#171) - address fatal errors
                        // in this case the binary representation of the message is invalid
273
                        panic!("fatal error - failed to decode message from stream; invalid line protocol: {e:?}");
274
275
                    }
                    None => {
276
                        tracing::debug!("tcp stream closed by server");
277
278
279
280
281
                        // If no Stop/Kill was received, this is a cancellation where frontend
                        // dropped the connection
                        if let Some(counter) = &cancellation_counter && !cancellation_counted {
                            counter.inc();
                        }
282
                        break;
283
284
285
286
287
288
289
290
                    }
                }
            }
            _ = alive_tx.closed() => {
                break;
            }
        }
    }
291
    framed_reader
292
293
}

294
295
async fn handle_writer(
    mut framed_writer: FramedWrite<tokio::io::WriteHalf<tokio::net::TcpStream>, TwoPartCodec>,
296
297
    mut bytes_rx: tokio::sync::mpsc::Receiver<TwoPartMessage>,
    alive_rx: tokio::sync::oneshot::Receiver<()>,
298
299
    context: Arc<dyn AsyncEngineContext>,
) -> Result<FramedWrite<tokio::io::WriteHalf<tokio::net::TcpStream>, TwoPartCodec>> {
300
301
302
    // Only send sentinel for normal channel closure
    let mut send_sentinel = true;

303
304
305
306
307
308
    loop {
        let msg = tokio::select! {
            biased;

            _ = context.killed() => {
                tracing::trace!("context kill signal received; shutting down");
309
                send_sentinel = false;
310
311
312
                break;
            }

313
314
            _ = context.stopped() => {
                tracing::trace!("context stop signal received; shutting down");
315
                send_sentinel = false;
316
317
318
                break;
            }

319
320
321
322
323
324
325
326
327
328
329
            msg = bytes_rx.recv() => {
                match msg {
                    Some(msg) => msg,
                    None => {
                        tracing::trace!("response channel closed; shutting down");
                        break;
                    }
                }
            }
        };

330
        if let Err(e) = framed_writer.send(msg).await {
331
            tracing::trace!(
332
                "failed to send message to network; possible disconnect: {:?}",
333
334
                e
            );
335
            send_sentinel = false;
336
337
338
            break;
        }
    }
339

340
341
342
343
344
345
    // Send sentinel only on normal closure
    if send_sentinel {
        let message = serde_json::to_vec(&ControlMessage::Sentinel)?;
        let msg = TwoPartMessage::from_header(message.into());
        framed_writer.send(msg).await?;
    }
346
347
348

    drop(alive_rx);
    Ok(framed_writer)
349
}
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744

#[cfg(test)]
mod tests {
    use super::*;
    use crate::pipeline::context::Controller;
    use crate::pipeline::network::tcp::test_utils::create_tcp_pair;
    use bytes::Bytes;
    use futures::StreamExt;
    use std::sync::Arc;
    use tokio::io::AsyncReadExt;
    use tokio::net::TcpStream;
    use tokio::sync::{mpsc, oneshot};
    use tokio_util::codec::FramedRead;

    struct WriterHarness {
        server: tokio::net::TcpStream,
        framed_writer: FramedWrite<tokio::io::WriteHalf<tokio::net::TcpStream>, TwoPartCodec>,
        bytes_tx: mpsc::Sender<TwoPartMessage>,
        bytes_rx: mpsc::Receiver<TwoPartMessage>,
        alive_tx: oneshot::Sender<()>,
        alive_rx: oneshot::Receiver<()>,
        controller: Arc<Controller>,
    }

    /// Creates a reusable writer harness with paired TCP streams and test channels.
    async fn writer_harness() -> WriterHarness {
        let (client, server) = create_tcp_pair().await;
        let (_, write_half) = tokio::io::split(client);
        let framed_writer = FramedWrite::new(write_half, TwoPartCodec::default());

        let (bytes_tx, bytes_rx) = mpsc::channel(64);
        let (alive_tx, alive_rx) = oneshot::channel::<()>();
        let controller = Arc::new(Controller::default());

        WriterHarness {
            server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_tx,
            alive_rx,
            controller,
        }
    }

    async fn recv_msg(reader: &mut FramedRead<TcpStream, TwoPartCodec>) -> TwoPartMessage {
        reader
            .next()
            .await
            .expect("expected message")
            .expect("failed to decode message")
    }

    fn assert_data_only_message(msg: TwoPartMessage, expected: &[u8]) {
        let (header, data) = msg.optional_parts();
        assert!(header.is_none(), "data-only message should not have header");
        assert_eq!(
            data.expect("data payload missing").as_ref(),
            expected,
            "data payload should match"
        );
    }

    fn assert_header_only_message(msg: TwoPartMessage, expected: &[u8]) {
        let (header, data) = msg.optional_parts();
        assert!(data.is_none(), "header-only message should not carry data");
        assert_eq!(
            header.expect("header missing").as_ref(),
            expected,
            "header payload should match"
        );
    }

    fn assert_header_and_data_message(
        msg: TwoPartMessage,
        expected_header: &[u8],
        expected_data: &[u8],
    ) {
        let (header, data) = msg.optional_parts();
        assert_eq!(
            header.expect("header missing").as_ref(),
            expected_header,
            "header payload should match"
        );
        assert_eq!(
            data.expect("data missing").as_ref(),
            expected_data,
            "data payload should match"
        );
    }

    fn assert_sentinel_message(msg: TwoPartMessage) {
        let (header, data) = msg.optional_parts();
        assert!(data.is_none(), "sentinel should not include a data section");
        let expected_sentinel = serde_json::to_vec(&ControlMessage::Sentinel).unwrap();
        assert_eq!(
            header.expect("sentinel header missing").as_ref(),
            expected_sentinel.as_slice(),
            "sentinel header should match serialized ControlMessage::Sentinel"
        );
    }

    /// Test that handle_writer forwards messages from the channel to the framed writer
    #[tokio::test]
    async fn test_handle_writer_forwards_messages() {
        let WriterHarness {
            server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Send test messages
        let test_msg = TwoPartMessage::from_data(Bytes::from("test data"));
        bytes_tx.send(test_msg).await.unwrap();

        // Close the sender to trigger normal termination
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // Decode from server side to verify data and sentinel were sent
        let mut reader = FramedRead::new(server, TwoPartCodec::default());

        let msg = recv_msg(&mut reader).await;
        assert_data_only_message(msg, b"test data");

        let sentinel = recv_msg(&mut reader).await;
        assert_sentinel_message(sentinel);
    }

    /// Test that handle_writer sends sentinel on normal channel closure
    #[tokio::test]
    async fn test_handle_writer_sends_sentinel_on_normal_closure() {
        let WriterHarness {
            mut server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Close the sender immediately to trigger normal termination
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // Read from server side to verify sentinel was sent
        let mut buffer = vec![0u8; 1024];
        let n = server.read(&mut buffer).await.unwrap();

        // Buffer should contain the sentinel message
        assert!(n > 0, "Expected sentinel to be written to the TCP stream");

        // Verify it contains the sentinel message by checking for the JSON
        let sentinel_json = serde_json::to_vec(&ControlMessage::Sentinel).unwrap();
        assert!(
            buffer[..n]
                .windows(sentinel_json.len())
                .any(|w| w == sentinel_json.as_slice()),
            "Buffer should contain sentinel message. Buffer: {:?}",
            String::from_utf8_lossy(&buffer[..n])
        );
    }

    /// Test that handle_writer does NOT send sentinel when context is killed
    #[tokio::test]
    async fn test_handle_writer_no_sentinel_on_context_killed() {
        let WriterHarness {
            mut server,
            framed_writer,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Kill the context
        controller.kill();

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // Drop the writer to close the connection, then try to read. Otherwise,
        // the test will hang on `server.read()`
        drop(result);

        // Read from server side - should get no sentinel
        let mut buffer = vec![0u8; 1024];
        let n = server.read(&mut buffer).await.unwrap();

        // Buffer should be empty (no sentinel sent)
        let sentinel_json = serde_json::to_vec(&ControlMessage::Sentinel).unwrap();
        assert!(
            n == 0
                || !buffer[..n]
                    .windows(sentinel_json.len())
                    .any(|w| w == sentinel_json.as_slice()),
            "Buffer should NOT contain sentinel message when context is killed"
        );
    }

    /// Test that handle_writer does NOT send sentinel when context is stopped
    #[tokio::test]
    async fn test_handle_writer_no_sentinel_on_context_stopped() {
        let WriterHarness {
            mut server,
            framed_writer,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Stop the context
        controller.stop();

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // Drop the writer to close the connection, then try to read. Otherwise,
        // the test will hang on `server.read()`
        drop(result);

        // Read from server side - should get no sentinel
        let mut buffer = vec![0u8; 1024];
        let n = server.read(&mut buffer).await.unwrap();

        // Buffer should be empty (no sentinel sent)
        let sentinel_json = serde_json::to_vec(&ControlMessage::Sentinel).unwrap();
        assert!(
            n == 0
                || !buffer[..n]
                    .windows(sentinel_json.len())
                    .any(|w| w == sentinel_json.as_slice()),
            "Buffer should NOT contain sentinel message when context is stopped"
        );
    }

    /// Test that handle_writer handles multiple messages correctly
    #[tokio::test]
    async fn test_handle_writer_multiple_messages() {
        let WriterHarness {
            server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Send multiple messages
        for i in 0..5 {
            let test_msg = TwoPartMessage::from_data(Bytes::from(format!("message {}", i)));
            bytes_tx.send(test_msg).await.unwrap();
        }

        // Close the sender to trigger normal termination
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // Decode from server side to verify all messages plus sentinel
        let mut reader = FramedRead::new(server, TwoPartCodec::default());
        for i in 0..5 {
            let msg = recv_msg(&mut reader).await;
            assert_data_only_message(msg, format!("message {}", i).as_bytes());
        }

        let sentinel = recv_msg(&mut reader).await;
        assert_sentinel_message(sentinel);
    }

    /// Test that alive_rx is dropped after handle_writer completes
    #[tokio::test]
    async fn test_handle_writer_drops_alive_rx() {
        let WriterHarness {
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_tx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Close the sender to trigger normal termination
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        // alive_tx should now be closed because alive_rx was dropped
        assert!(alive_tx.is_closed());
    }

    /// Test handle_writer with header-only messages (control messages)
    #[tokio::test]
    async fn test_handle_writer_header_only_messages() {
        let WriterHarness {
            server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Send a header-only message
        let header_msg = TwoPartMessage::from_header(Bytes::from("header content"));
        bytes_tx.send(header_msg).await.unwrap();

        // Close the sender
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        let mut reader = FramedRead::new(server, TwoPartCodec::default());

        let header_msg = recv_msg(&mut reader).await;
        assert_header_only_message(header_msg, b"header content");

        let sentinel = recv_msg(&mut reader).await;
        assert_sentinel_message(sentinel);
    }

    /// Test handle_writer with mixed header and data messages
    #[tokio::test]
    async fn test_handle_writer_mixed_messages() {
        let WriterHarness {
            server,
            framed_writer,
            bytes_tx,
            bytes_rx,
            alive_rx,
            controller,
            ..
        } = writer_harness().await;

        // Send mixed messages
        bytes_tx
            .send(TwoPartMessage::from_header(Bytes::from("header1")))
            .await
            .unwrap();
        bytes_tx
            .send(TwoPartMessage::from_data(Bytes::from("data1")))
            .await
            .unwrap();
        bytes_tx
            .send(TwoPartMessage::from_parts(
                Bytes::from("header2"),
                Bytes::from("data2"),
            ))
            .await
            .unwrap();

        // Close the sender
        drop(bytes_tx);

        let result = handle_writer(framed_writer, bytes_rx, alive_rx, controller).await;

        assert!(result.is_ok());

        let mut reader = FramedRead::new(server, TwoPartCodec::default());

        let first = recv_msg(&mut reader).await;
        assert_header_only_message(first, b"header1");

        let second = recv_msg(&mut reader).await;
        assert_data_only_message(second, b"data1");

        let third = recv_msg(&mut reader).await;
        assert_header_and_data_message(third, b"header2", b"data2");

        let sentinel = recv_msg(&mut reader).await;
        assert_sentinel_message(sentinel);
    }
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793

    // ==================== handle_reader tests ====================

    struct ReaderHarness {
        framed_server: FramedWrite<tokio::io::WriteHalf<tokio::net::TcpStream>, TwoPartCodec>,
        framed_reader: FramedRead<tokio::io::ReadHalf<tokio::net::TcpStream>, TwoPartCodec>,
        alive_tx: oneshot::Sender<()>,
        alive_rx: oneshot::Receiver<()>,
        controller: Arc<Controller>,
    }

    /// Creates a reusable reader harness with paired TCP streams and test channels.
    async fn reader_harness() -> ReaderHarness {
        let (client, server) = create_tcp_pair().await;
        let (read_half, _write_half) = tokio::io::split(client);
        let (_server_read, server_write) = tokio::io::split(server);

        let framed_reader = FramedRead::new(read_half, TwoPartCodec::default());
        let framed_server = FramedWrite::new(server_write, TwoPartCodec::default());
        let (alive_tx, alive_rx) = oneshot::channel::<()>();
        let controller = Arc::new(Controller::default());

        ReaderHarness {
            framed_server,
            framed_reader,
            alive_tx,
            alive_rx,
            controller,
        }
    }

    fn control_message(msg: &ControlMessage) -> TwoPartMessage {
        let msg_bytes = serde_json::to_vec(msg).unwrap();
        TwoPartMessage::from_header(Bytes::from(msg_bytes))
    }

    /// Test that handle_reader handles Stop control message by calling context.stop()
    #[tokio::test]
    async fn test_handle_reader_stop_control_message() {
        let ReaderHarness {
            mut framed_server,
            framed_reader,
            alive_tx,
            alive_rx: _alive_rx,
            controller,
        } = reader_harness().await;

        // Spawn the reader task
        let controller_clone = controller.clone();
794
795
796
        let reader_handle = tokio::spawn(async move {
            handle_reader(framed_reader, controller_clone, alive_tx, None).await
        });
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829

        // Send Stop control message from server
        framed_server
            .send(control_message(&ControlMessage::Stop))
            .await
            .unwrap();

        // Close the framed server to signal EOF to the client
        framed_server.close().await.unwrap();

        // Wait for reader to finish
        let _ = reader_handle.await.unwrap();

        // Verify that stop was called on the controller
        assert!(
            controller.is_stopped(),
            "Controller should be stopped after receiving Stop message"
        );
    }

    /// Test that handle_reader handles Kill control message by calling context.kill()
    #[tokio::test]
    async fn test_handle_reader_kill_control_message() {
        let ReaderHarness {
            mut framed_server,
            framed_reader,
            alive_tx,
            alive_rx: _alive_rx,
            controller,
        } = reader_harness().await;

        // Spawn the reader task
        let controller_clone = controller.clone();
830
831
832
        let reader_handle = tokio::spawn(async move {
            handle_reader(framed_reader, controller_clone, alive_tx, None).await
        });
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865

        // Send Kill control message from server
        framed_server
            .send(control_message(&ControlMessage::Kill))
            .await
            .unwrap();

        // Close the framed server to signal EOF to the client
        framed_server.close().await.unwrap();

        // Wait for reader to finish
        let _ = reader_handle.await.unwrap();

        // Verify that kill was called on the controller
        assert!(
            controller.is_killed(),
            "Controller should be killed after receiving Kill message"
        );
    }

    /// Test that handle_reader exits when alive channel is closed
    #[tokio::test]
    async fn test_handle_reader_exits_on_alive_channel_closed() {
        let ReaderHarness {
            framed_reader,
            alive_tx,
            alive_rx,
            controller,
            ..
        } = reader_harness().await;

        // Spawn the reader task
        let reader_handle =
866
867
868
            tokio::spawn(
                async move { handle_reader(framed_reader, controller, alive_tx, None).await },
            );
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894

        // Drop the alive_rx to close the channel (simulating writer finishing)
        drop(alive_rx);

        // Reader should exit due to alive channel closure
        let result = reader_handle.await;

        assert!(
            result.is_ok(),
            "handle_reader should exit when alive channel is closed"
        );
    }

    /// Test that handle_reader exits when TCP stream is closed
    #[tokio::test]
    async fn test_handle_reader_exits_on_stream_closed() {
        let ReaderHarness {
            mut framed_server,
            framed_reader,
            alive_tx,
            alive_rx: _alive_rx,
            controller,
        } = reader_harness().await;

        // Spawn the reader task
        let reader_handle =
895
896
897
            tokio::spawn(
                async move { handle_reader(framed_reader, controller, alive_tx, None).await },
            );
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923

        // Close the framed server to signal EOF to the client
        framed_server.close().await.unwrap();

        // Reader should exit due to stream closure
        let result = tokio::time::timeout(std::time::Duration::from_secs(1), reader_handle).await;

        assert!(
            result.is_ok(),
            "handle_reader should exit when stream is closed"
        );
    }

    /// Test that handle_reader handles multiple control messages in sequence
    #[tokio::test]
    async fn test_handle_reader_multiple_control_messages() {
        let ReaderHarness {
            mut framed_server,
            framed_reader,
            alive_tx,
            alive_rx: _alive_rx,
            controller,
        } = reader_harness().await;

        // Spawn the reader task
        let controller_clone = controller.clone();
924
925
926
        let reader_handle = tokio::spawn(async move {
            handle_reader(framed_reader, controller_clone, alive_tx, None).await
        });
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963

        // Send multiple Stop messages (first one will stop, subsequent ones are no-ops)
        framed_server
            .send(control_message(&ControlMessage::Stop))
            .await
            .unwrap();
        framed_server
            .send(control_message(&ControlMessage::Stop))
            .await
            .unwrap();

        // Close the framed server to signal EOF to the client
        framed_server.close().await.unwrap();

        // Wait for reader to finish
        let _ = reader_handle.await.unwrap();

        // Verify that stop was called
        assert!(
            controller.is_stopped(),
            "Controller should be stopped after receiving Stop messages"
        );
    }

    /// Test handle_reader with Stop followed by Kill
    #[tokio::test]
    async fn test_handle_reader_stop_then_kill() {
        let ReaderHarness {
            mut framed_server,
            framed_reader,
            alive_tx,
            alive_rx: _alive_rx,
            controller,
        } = reader_harness().await;

        // Spawn the reader task
        let controller_clone = controller.clone();
964
965
966
        let reader_handle = tokio::spawn(async move {
            handle_reader(framed_reader, controller_clone, alive_tx, None).await
        });
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989

        // Send Stop first, then Kill
        framed_server
            .send(control_message(&ControlMessage::Stop))
            .await
            .unwrap();
        framed_server
            .send(control_message(&ControlMessage::Kill))
            .await
            .unwrap();

        // Close the framed server to signal EOF to the client
        framed_server.close().await.unwrap();

        // Wait for reader to finish
        let _ = reader_handle.await.unwrap();

        // Verify that kill was called (which sets killed state)
        assert!(
            controller.is_killed(),
            "Controller should be killed after receiving Kill message"
        );
    }
990
}