risingwave_meta/barrier/
complete_task.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::collections::HashMap;
16use std::future::{Future, pending};
17use std::mem::replace;
18use std::sync::Arc;
19
20use anyhow::Context;
21use futures::future::try_join_all;
22use prometheus::HistogramTimer;
23use risingwave_common::catalog::{DatabaseId, TableId};
24use risingwave_common::must_match;
25use risingwave_common::util::deployment::Deployment;
26use risingwave_pb::hummock::HummockVersionStats;
27use tokio::task::JoinHandle;
28
29use crate::barrier::checkpoint::CheckpointControl;
30use crate::barrier::command::CommandContext;
31use crate::barrier::context::GlobalBarrierWorkerContext;
32use crate::barrier::notifier::Notifier;
33use crate::barrier::progress::TrackingJob;
34use crate::barrier::rpc::ControlStreamManager;
35use crate::barrier::schedule::PeriodicBarriers;
36use crate::hummock::CommitEpochInfo;
37use crate::manager::MetaSrvEnv;
38use crate::rpc::metrics::GLOBAL_META_METRICS;
39use crate::{MetaError, MetaResult};
40
41pub(super) enum CompletingTask {
42    None,
43    Completing {
44        #[expect(clippy::type_complexity)]
45        /// `database_id` -> (`Some(database_graph_committed_epoch)`, [(`creating_job_id`, `creating_job_committed_epoch`)])
46        epochs_to_ack: HashMap<DatabaseId, (Option<u64>, Vec<(TableId, u64)>)>,
47
48        // The join handle of a spawned task that completes the barrier.
49        // The return value indicate whether there is some create streaming job command
50        // that has finished but not checkpointed. If there is any, we will force checkpoint on the next barrier
51        join_handle: JoinHandle<MetaResult<HummockVersionStats>>,
52    },
53    #[expect(dead_code)]
54    Err(MetaError),
55}
56
57#[derive(Default)]
58pub(super) struct CompleteBarrierTask {
59    pub(super) commit_info: CommitEpochInfo,
60    pub(super) finished_jobs: Vec<TrackingJob>,
61    pub(super) notifiers: Vec<Notifier>,
62    /// `database_id` -> (Some((`command_ctx`, `enqueue_time`)), vec!((`creating_job_id`, `epoch`)))
63    #[expect(clippy::type_complexity)]
64    pub(super) epoch_infos: HashMap<
65        DatabaseId,
66        (
67            Option<(CommandContext, HistogramTimer)>,
68            Vec<(TableId, u64)>,
69        ),
70    >,
71}
72
73impl CompleteBarrierTask {
74    #[expect(clippy::type_complexity)]
75    pub(super) fn epochs_to_ack(&self) -> HashMap<DatabaseId, (Option<u64>, Vec<(TableId, u64)>)> {
76        self.epoch_infos
77            .iter()
78            .map(|(database_id, (command_context, creating_job_epochs))| {
79                (
80                    *database_id,
81                    (
82                        command_context
83                            .as_ref()
84                            .map(|(command, _)| command.barrier_info.prev_epoch.value().0),
85                        creating_job_epochs.clone(),
86                    ),
87                )
88            })
89            .collect()
90    }
91}
92
93impl CompleteBarrierTask {
94    pub(super) async fn complete_barrier(
95        self,
96        context: &impl GlobalBarrierWorkerContext,
97        env: MetaSrvEnv,
98    ) -> MetaResult<HummockVersionStats> {
99        let result: MetaResult<HummockVersionStats> = try {
100            let wait_commit_timer = GLOBAL_META_METRICS
101                .barrier_wait_commit_latency
102                .start_timer();
103            let version_stats = context.commit_epoch(self.commit_info).await?;
104            for command_ctx in self
105                .epoch_infos
106                .values()
107                .flat_map(|(command, _)| command.as_ref().map(|(command, _)| command))
108            {
109                context.post_collect_command(command_ctx).await?;
110            }
111
112            wait_commit_timer.observe_duration();
113            version_stats
114        };
115
116        let version_stats = {
117            let version_stats = match result {
118                Ok(version_stats) => version_stats,
119                Err(e) => {
120                    for notifier in self.notifiers {
121                        notifier.notify_collection_failed(e.clone());
122                    }
123                    return Err(e);
124                }
125            };
126            self.notifiers.into_iter().for_each(|notifier| {
127                notifier.notify_collected();
128            });
129            try_join_all(
130                self.finished_jobs
131                    .into_iter()
132                    .map(|finished_job| context.finish_creating_job(finished_job)),
133            )
134            .await?;
135            for (database_id, (command, _)) in self.epoch_infos {
136                if let Some((command_ctx, enqueue_time)) = command {
137                    let duration_sec = enqueue_time.stop_and_record();
138                    Self::report_complete_event(&env, duration_sec, &command_ctx);
139                    GLOBAL_META_METRICS
140                        .last_committed_barrier_time
141                        .with_label_values(&[database_id.database_id.to_string().as_str()])
142                        .set(command_ctx.barrier_info.curr_epoch.value().as_unix_secs() as i64);
143                }
144            }
145            version_stats
146        };
147
148        Ok(version_stats)
149    }
150}
151
152impl CompleteBarrierTask {
153    fn report_complete_event(env: &MetaSrvEnv, duration_sec: f64, command_ctx: &CommandContext) {
154        // Record barrier latency in event log.
155        use risingwave_pb::meta::event_log;
156        let event = event_log::EventBarrierComplete {
157            prev_epoch: command_ctx.barrier_info.prev_epoch(),
158            cur_epoch: command_ctx.barrier_info.curr_epoch.value().0,
159            duration_sec,
160            command: command_ctx
161                .command
162                .as_ref()
163                .map(|command| command.to_string())
164                .unwrap_or_else(|| "barrier".to_owned()),
165            barrier_kind: command_ctx.barrier_info.kind.as_str_name().to_owned(),
166        };
167        if cfg!(debug_assertions) || Deployment::current().is_ci() {
168            // Add a warning log so that debug mode / CI can observe it
169            if duration_sec > 5.0 {
170                tracing::warn!(event = ?event,"high barrier latency observed!")
171            }
172        }
173        env.event_log_manager_ref()
174            .add_event_logs(vec![event_log::Event::BarrierComplete(event)]);
175    }
176}
177
178pub(super) struct BarrierCompleteOutput {
179    #[expect(clippy::type_complexity)]
180    /// `database_id` -> (`Some(database_graph_committed_epoch)`, [(`creating_job_id`, `creating_job_committed_epoch`)])
181    pub epochs_to_ack: HashMap<DatabaseId, (Option<u64>, Vec<(TableId, u64)>)>,
182    pub hummock_version_stats: HummockVersionStats,
183}
184
185impl CompletingTask {
186    pub(super) fn next_completed_barrier<'a>(
187        &'a mut self,
188        periodic_barriers: &mut PeriodicBarriers,
189        checkpoint_control: &mut CheckpointControl,
190        control_stream_manager: &mut ControlStreamManager,
191        context: &Arc<impl GlobalBarrierWorkerContext>,
192        env: &MetaSrvEnv,
193    ) -> impl Future<Output = MetaResult<BarrierCompleteOutput>> + 'a {
194        // If there is no completing barrier, try to start completing the earliest barrier if
195        // it has been collected.
196        if let CompletingTask::None = self {
197            if let Some(task) = checkpoint_control
198                .next_complete_barrier_task(Some((periodic_barriers, control_stream_manager)))
199            {
200                {
201                    let epochs_to_ack = task.epochs_to_ack();
202                    let context = context.clone();
203                    let await_tree_reg = env.await_tree_reg().clone();
204                    let env = env.clone();
205
206                    let fut = async move { task.complete_barrier(&*context, env).await };
207                    let fut = await_tree_reg
208                        .register_derived_root("Barrier Completion Task")
209                        .instrument(fut);
210                    let join_handle = tokio::spawn(fut);
211
212                    *self = CompletingTask::Completing {
213                        epochs_to_ack,
214                        join_handle,
215                    };
216                }
217            }
218        }
219
220        async move {
221            if !matches!(self, CompletingTask::Completing { .. }) {
222                return pending().await;
223            };
224            self.next_completed_barrier_inner().await
225        }
226    }
227
228    #[await_tree::instrument]
229    pub(super) async fn wait_completing_task(
230        &mut self,
231    ) -> MetaResult<Option<BarrierCompleteOutput>> {
232        match self {
233            CompletingTask::None => Ok(None),
234            CompletingTask::Completing { .. } => {
235                self.next_completed_barrier_inner().await.map(Some)
236            }
237            CompletingTask::Err(_) => {
238                unreachable!("should not be called on previous err")
239            }
240        }
241    }
242
243    async fn next_completed_barrier_inner(&mut self) -> MetaResult<BarrierCompleteOutput> {
244        let CompletingTask::Completing { join_handle, .. } = self else {
245            unreachable!()
246        };
247
248        {
249            {
250                let join_result: MetaResult<_> = try {
251                    join_handle
252                        .await
253                        .context("failed to join completing command")??
254                };
255                // It's important to reset the completing_command after await no matter the result is err
256                // or not, and otherwise the join handle will be polled again after ready.
257                let next_completing_command_status = if let Err(e) = &join_result {
258                    CompletingTask::Err(e.clone())
259                } else {
260                    CompletingTask::None
261                };
262                let completed_command = replace(self, next_completing_command_status);
263                let hummock_version_stats = join_result?;
264
265                must_match!(completed_command, CompletingTask::Completing {
266                    epochs_to_ack,
267                    ..
268                } => {
269                    Ok(BarrierCompleteOutput {
270                        epochs_to_ack,
271                        hummock_version_stats,
272                    })
273                })
274            }
275        }
276    }
277}
risingwave_meta/barrier/complete_task.rs

risingwave_meta/barrier/
complete_task.rs
