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/// Only for checkpoint barrier. For normal barrier, there won't be a task.
58#[derive(Default)]
59pub(super) struct CompleteBarrierTask {
60    pub(super) commit_info: CommitEpochInfo,
61    pub(super) finished_jobs: Vec<TrackingJob>,
62    pub(super) notifiers: Vec<Notifier>,
63    /// `database_id` -> (Some((`command_ctx`, `enqueue_time`)), vec!((`creating_job_id`, `epoch`)))
64    #[expect(clippy::type_complexity)]
65    pub(super) epoch_infos: HashMap<
66        DatabaseId,
67        (
68            Option<(CommandContext, HistogramTimer)>,
69            Vec<(TableId, u64)>,
70        ),
71    >,
72    /// Source IDs that have finished loading data and need `LoadFinish` commands
73    pub(super) load_finished_source_ids: Vec<u32>,
74}
75
76impl CompleteBarrierTask {
77    #[expect(clippy::type_complexity)]
78    pub(super) fn epochs_to_ack(&self) -> HashMap<DatabaseId, (Option<u64>, Vec<(TableId, u64)>)> {
79        self.epoch_infos
80            .iter()
81            .map(|(database_id, (command_context, creating_job_epochs))| {
82                (
83                    *database_id,
84                    (
85                        command_context
86                            .as_ref()
87                            .map(|(command, _)| command.barrier_info.prev_epoch.value().0),
88                        creating_job_epochs.clone(),
89                    ),
90                )
91            })
92            .collect()
93    }
94}
95
96impl CompleteBarrierTask {
97    pub(super) async fn complete_barrier(
98        self,
99        context: &impl GlobalBarrierWorkerContext,
100        env: MetaSrvEnv,
101    ) -> MetaResult<HummockVersionStats> {
102        let result: MetaResult<HummockVersionStats> = try {
103            let wait_commit_timer = GLOBAL_META_METRICS
104                .barrier_wait_commit_latency
105                .start_timer();
106            let version_stats = context.commit_epoch(self.commit_info).await?;
107
108            // Handle load finished source IDs for refreshable batch sources
109            // Spawn this asynchronously to avoid deadlock during barrier collection
110            if !self.load_finished_source_ids.is_empty() {
111                context
112                    .handle_load_finished_source_ids(self.load_finished_source_ids.clone())
113                    .await?;
114            }
115
116            for command_ctx in self
117                .epoch_infos
118                .values()
119                .flat_map(|(command, _)| command.as_ref().map(|(command, _)| command))
120            {
121                context.post_collect_command(command_ctx).await?;
122            }
123
124            wait_commit_timer.observe_duration();
125            version_stats
126        };
127
128        let version_stats = {
129            let version_stats = match result {
130                Ok(version_stats) => version_stats,
131                Err(e) => {
132                    for notifier in self.notifiers {
133                        notifier.notify_collection_failed(e.clone());
134                    }
135                    return Err(e);
136                }
137            };
138            self.notifiers.into_iter().for_each(|notifier| {
139                notifier.notify_collected();
140            });
141            try_join_all(
142                self.finished_jobs
143                    .into_iter()
144                    .map(|finished_job| context.finish_creating_job(finished_job)),
145            )
146            .await?;
147            for (database_id, (command, _)) in self.epoch_infos {
148                if let Some((command_ctx, enqueue_time)) = command {
149                    let duration_sec = enqueue_time.stop_and_record();
150                    Self::report_complete_event(&env, duration_sec, &command_ctx);
151                    GLOBAL_META_METRICS
152                        .last_committed_barrier_time
153                        .with_label_values(&[database_id.database_id.to_string().as_str()])
154                        .set(command_ctx.barrier_info.curr_epoch.value().as_unix_secs() as i64);
155                }
156            }
157            version_stats
158        };
159
160        Ok(version_stats)
161    }
162}
163
164impl CompleteBarrierTask {
165    fn report_complete_event(env: &MetaSrvEnv, duration_sec: f64, command_ctx: &CommandContext) {
166        // Record barrier latency in event log.
167        use risingwave_pb::meta::event_log;
168        let event = event_log::EventBarrierComplete {
169            prev_epoch: command_ctx.barrier_info.prev_epoch(),
170            cur_epoch: command_ctx.barrier_info.curr_epoch.value().0,
171            duration_sec,
172            command: command_ctx
173                .command
174                .as_ref()
175                .map(|command| command.to_string())
176                .unwrap_or_else(|| "barrier".to_owned()),
177            barrier_kind: command_ctx.barrier_info.kind.as_str_name().to_owned(),
178        };
179        if cfg!(debug_assertions) || Deployment::current().is_ci() {
180            // Add a warning log so that debug mode / CI can observe it
181            if duration_sec > 5.0 {
182                tracing::warn!(event = ?event,"high barrier latency observed!")
183            }
184        }
185        env.event_log_manager_ref()
186            .add_event_logs(vec![event_log::Event::BarrierComplete(event)]);
187    }
188}
189
190pub(super) struct BarrierCompleteOutput {
191    #[expect(clippy::type_complexity)]
192    /// `database_id` -> (`Some(database_graph_committed_epoch)`, [(`creating_job_id`, `creating_job_committed_epoch`)])
193    pub epochs_to_ack: HashMap<DatabaseId, (Option<u64>, Vec<(TableId, u64)>)>,
194    pub hummock_version_stats: HummockVersionStats,
195}
196
197impl CompletingTask {
198    pub(super) fn next_completed_barrier<'a>(
199        &'a mut self,
200        periodic_barriers: &mut PeriodicBarriers,
201        checkpoint_control: &mut CheckpointControl,
202        control_stream_manager: &mut ControlStreamManager,
203        context: &Arc<impl GlobalBarrierWorkerContext>,
204        env: &MetaSrvEnv,
205    ) -> impl Future<Output = MetaResult<BarrierCompleteOutput>> + 'a {
206        // If there is no completing barrier, try to start completing the earliest barrier if
207        // it has been collected.
208        if let CompletingTask::None = self
209            && let Some(task) = checkpoint_control
210                .next_complete_barrier_task(Some((periodic_barriers, control_stream_manager)))
211        {
212            {
213                let epochs_to_ack = task.epochs_to_ack();
214                let context = context.clone();
215                let await_tree_reg = env.await_tree_reg().clone();
216                let env = env.clone();
217
218                let fut = async move { task.complete_barrier(&*context, env).await };
219                let fut = await_tree_reg
220                    .register_derived_root("Barrier Completion Task")
221                    .instrument(fut);
222                let join_handle = tokio::spawn(fut);
223
224                *self = CompletingTask::Completing {
225                    epochs_to_ack,
226                    join_handle,
227                };
228            }
229        }
230
231        async move {
232            if !matches!(self, CompletingTask::Completing { .. }) {
233                return pending().await;
234            };
235            self.next_completed_barrier_inner().await
236        }
237    }
238
239    #[await_tree::instrument]
240    pub(super) async fn wait_completing_task(
241        &mut self,
242    ) -> MetaResult<Option<BarrierCompleteOutput>> {
243        match self {
244            CompletingTask::None => Ok(None),
245            CompletingTask::Completing { .. } => {
246                self.next_completed_barrier_inner().await.map(Some)
247            }
248            CompletingTask::Err(_) => {
249                unreachable!("should not be called on previous err")
250            }
251        }
252    }
253
254    async fn next_completed_barrier_inner(&mut self) -> MetaResult<BarrierCompleteOutput> {
255        let CompletingTask::Completing { join_handle, .. } = self else {
256            unreachable!()
257        };
258
259        {
260            {
261                let join_result: MetaResult<_> = try {
262                    join_handle
263                        .await
264                        .context("failed to join completing command")??
265                };
266                // It's important to reset the completing_command after await no matter the result is err
267                // or not, and otherwise the join handle will be polled again after ready.
268                let next_completing_command_status = if let Err(e) = &join_result {
269                    CompletingTask::Err(e.clone())
270                } else {
271                    CompletingTask::None
272                };
273                let completed_command = replace(self, next_completing_command_status);
274                let hummock_version_stats = join_result?;
275
276                must_match!(completed_command, CompletingTask::Completing {
277                    epochs_to_ack,
278                    ..
279                } => {
280                    Ok(BarrierCompleteOutput {
281                        epochs_to_ack,
282                        hummock_version_stats,
283                    })
284                })
285            }
286        }
287    }
288}
risingwave_meta/barrier/complete_task.rs

risingwave_meta/barrier/
complete_task.rs
