risingwave_compute/rpc/service/
exchange_service.rs

1// Copyright 2025 RisingWave Labs
2//
3// Licensed under the Apache License, Version 2.0 (the "License");
4// you may not use this file except in compliance with the License.
5// You may obtain a copy of the License at
6//
7//     http://www.apache.org/licenses/LICENSE-2.0
8//
9// Unless required by applicable law or agreed to in writing, software
10// distributed under the License is distributed on an "AS IS" BASIS,
11// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12// See the License for the specific language governing permissions and
13// limitations under the License.
14
15use std::net::SocketAddr;
16use std::sync::Arc;
17
18use either::Either;
19use futures::{Stream, StreamExt, TryStreamExt, pin_mut};
20use futures_async_stream::try_stream;
21use risingwave_batch::task::BatchManager;
22use risingwave_common::catalog::DatabaseId;
23use risingwave_pb::task_service::exchange_service_server::ExchangeService;
24use risingwave_pb::task_service::{
25    GetDataRequest, GetDataResponse, GetStreamRequest, GetStreamResponse, PbPermits, permits,
26};
27use risingwave_stream::executor::DispatcherMessageBatch;
28use risingwave_stream::executor::exchange::permit::{MessageWithPermits, Receiver};
29use risingwave_stream::task::LocalStreamManager;
30use thiserror_ext::AsReport;
31use tokio_stream::wrappers::ReceiverStream;
32use tonic::{Request, Response, Status, Streaming};
33
34use crate::rpc::service::exchange_metrics::ExchangeServiceMetrics;
35
36#[derive(Clone)]
37pub struct ExchangeServiceImpl {
38    batch_mgr: Arc<BatchManager>,
39    stream_mgr: LocalStreamManager,
40    metrics: Arc<ExchangeServiceMetrics>,
41}
42
43pub type BatchDataStream = ReceiverStream<std::result::Result<GetDataResponse, Status>>;
44pub type StreamDataStream = impl Stream<Item = std::result::Result<GetStreamResponse, Status>>;
45
46#[async_trait::async_trait]
47impl ExchangeService for ExchangeServiceImpl {
48    type GetDataStream = BatchDataStream;
49    type GetStreamStream = StreamDataStream;
50
51    async fn get_data(
52        &self,
53        request: Request<GetDataRequest>,
54    ) -> std::result::Result<Response<Self::GetDataStream>, Status> {
55        let peer_addr = request
56            .remote_addr()
57            .ok_or_else(|| Status::unavailable("connection unestablished"))?;
58        let pb_task_output_id = request
59            .into_inner()
60            .task_output_id
61            .expect("Failed to get task output id.");
62        let (tx, rx) =
63            tokio::sync::mpsc::channel(self.batch_mgr.config().developer.receiver_channel_size);
64        if let Err(e) = self.batch_mgr.get_data(tx, peer_addr, &pb_task_output_id) {
65            error!(
66                %peer_addr,
67                error = %e.as_report(),
68                "Failed to serve exchange RPC"
69            );
70            return Err(e.into());
71        }
72
73        Ok(Response::new(ReceiverStream::new(rx)))
74    }
75
76    async fn get_stream(
77        &self,
78        request: Request<Streaming<GetStreamRequest>>,
79    ) -> std::result::Result<Response<Self::GetStreamStream>, Status> {
80        use risingwave_pb::task_service::get_stream_request::*;
81
82        let peer_addr = request
83            .remote_addr()
84            .ok_or_else(|| Status::unavailable("get_stream connection unestablished"))?;
85
86        let mut request_stream: Streaming<GetStreamRequest> = request.into_inner();
87
88        // Extract the first `Get` request from the stream.
89        let Get {
90            up_actor_id,
91            down_actor_id,
92            up_fragment_id,
93            down_fragment_id,
94            database_id,
95            term_id,
96        } = {
97            let req = request_stream
98                .next()
99                .await
100                .ok_or_else(|| Status::invalid_argument("get_stream request is empty"))??;
101            match req.value.unwrap() {
102                Value::Get(get) => get,
103                Value::AddPermits(_) => unreachable!("the first message must be `Get`"),
104            }
105        };
106
107        let receiver = self
108            .stream_mgr
109            .take_receiver(
110                DatabaseId::new(database_id),
111                term_id,
112                (up_actor_id, down_actor_id),
113            )
114            .await?;
115
116        // Map the remaining stream to add-permits.
117        let add_permits_stream = request_stream.map_ok(|req| match req.value.unwrap() {
118            Value::Get(_) => unreachable!("the following messages must be `AddPermits`"),
119            Value::AddPermits(add_permits) => add_permits.value.unwrap(),
120        });
121
122        Ok(Response::new(Self::get_stream_impl(
123            self.metrics.clone(),
124            peer_addr,
125            receiver,
126            add_permits_stream,
127            (up_fragment_id, down_fragment_id),
128        )))
129    }
130}
131
132impl ExchangeServiceImpl {
133    pub fn new(
134        mgr: Arc<BatchManager>,
135        stream_mgr: LocalStreamManager,
136        metrics: Arc<ExchangeServiceMetrics>,
137    ) -> Self {
138        ExchangeServiceImpl {
139            batch_mgr: mgr,
140            stream_mgr,
141            metrics,
142        }
143    }
144
145    #[try_stream(ok = GetStreamResponse, error = Status)]
146    async fn get_stream_impl(
147        metrics: Arc<ExchangeServiceMetrics>,
148        peer_addr: SocketAddr,
149        mut receiver: Receiver,
150        add_permits_stream: impl Stream<Item = std::result::Result<permits::Value, tonic::Status>>,
151        up_down_fragment_ids: (u32, u32),
152    ) {
153        tracing::debug!(target: "events::compute::exchange", peer_addr = %peer_addr, "serve stream exchange RPC");
154        let up_fragment_id = up_down_fragment_ids.0.to_string();
155        let down_fragment_id = up_down_fragment_ids.1.to_string();
156
157        let permits = receiver.permits();
158
159        // Select from the permits back from the downstream and the upstream receiver.
160        let select_stream = futures::stream::select(
161            add_permits_stream.map_ok(Either::Left),
162            #[try_stream]
163            async move {
164                while let Some(m) = receiver.recv_raw().await {
165                    yield Either::Right(m);
166                }
167            },
168        );
169        pin_mut!(select_stream);
170
171        while let Some(r) = select_stream.try_next().await? {
172            match r {
173                Either::Left(permits_to_add) => {
174                    permits.add_permits(permits_to_add);
175                }
176                Either::Right(MessageWithPermits { message, permits }) => {
177                    let message = match message {
178                        DispatcherMessageBatch::Chunk(chunk) => {
179                            DispatcherMessageBatch::Chunk(chunk.compact())
180                        }
181                        msg @ (DispatcherMessageBatch::Watermark(_)
182                        | DispatcherMessageBatch::BarrierBatch(_)) => msg,
183                    };
184                    let proto = message.to_protobuf();
185                    // forward the acquired permit to the downstream
186                    let response = GetStreamResponse {
187                        message: Some(proto),
188                        permits: Some(PbPermits { value: permits }),
189                    };
190                    let bytes = DispatcherMessageBatch::get_encoded_len(&response);
191
192                    yield response;
193
194                    metrics
195                        .stream_fragment_exchange_bytes
196                        .with_label_values(&[&up_fragment_id, &down_fragment_id])
197                        .inc_by(bytes as u64);
198                }
199            }
200        }
201    }
202}