risingwave_connector/source/cdc/
mod.rs

1// Copyright 2022 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
15pub mod enumerator;
16pub mod external;
17pub mod jni_source;
18pub mod source;
19pub mod split;
20
21use std::collections::{BTreeMap, HashMap};
22use std::marker::PhantomData;
23
24pub use enumerator::*;
25use itertools::Itertools;
26use risingwave_common::id::{ActorId, SourceId};
27use risingwave_pb::catalog::PbSource;
28use risingwave_pb::connector_service::{PbSourceType, PbTableSchema, SourceType, TableSchema};
29use risingwave_pb::plan_common::ExternalTableDesc;
30use risingwave_pb::plan_common::column_desc::GeneratedOrDefaultColumn;
31use risingwave_pb::source::{PbCdcTableSnapshotSplit, PbCdcTableSnapshotSplitsWithGeneration};
32use risingwave_pb::stream_plan::StreamCdcScanOptions;
33use simd_json::prelude::ArrayTrait;
34pub use source::*;
35
36use crate::enforce_secret::EnforceSecret;
37use crate::error::ConnectorResult;
38use crate::source::{CdcTableSnapshotSplitRaw, SourceProperties, SplitImpl, TryFromBTreeMap};
39use crate::{for_all_classified_sources, impl_cdc_source_type};
40
41pub const CDC_CONNECTOR_NAME_SUFFIX: &str = "-cdc";
42pub const CDC_SNAPSHOT_MODE_KEY: &str = "debezium.snapshot.mode";
43pub const CDC_SNAPSHOT_BACKFILL: &str = "rw_cdc_backfill";
44pub const CDC_SHARING_MODE_KEY: &str = "rw.sharing.mode.enable";
45// User can set snapshot='false' to disable cdc backfill
46pub const CDC_BACKFILL_ENABLE_KEY: &str = "snapshot";
47pub const CDC_BACKFILL_SNAPSHOT_INTERVAL_KEY: &str = "snapshot.interval";
48pub const CDC_BACKFILL_SNAPSHOT_BATCH_SIZE_KEY: &str = "snapshot.batch_size";
49pub const CDC_BACKFILL_PARALLELISM: &str = "backfill.parallelism";
50pub const CDC_BACKFILL_NUM_ROWS_PER_SPLIT: &str = "backfill.num_rows_per_split";
51pub const CDC_BACKFILL_AS_EVEN_SPLITS: &str = "backfill.as_even_splits";
52pub const CDC_BACKFILL_SPLIT_PK_COLUMN_INDEX: &str = "backfill.split_pk_column_index";
53// We enable transaction for shared cdc source by default
54pub const CDC_TRANSACTIONAL_KEY: &str = "transactional";
55pub const CDC_WAIT_FOR_STREAMING_START_TIMEOUT: &str = "cdc.source.wait.streaming.start.timeout";
56pub const CDC_BACKFILL_MAX_PARALLELISM: u32 = 256;
57
58// User can set strong-schema='true' to enable strong schema for mongo cdc source
59pub const CDC_MONGODB_STRONG_SCHEMA_KEY: &str = "strong_schema";
60
61pub const MYSQL_CDC_CONNECTOR: &str = Mysql::CDC_CONNECTOR_NAME;
62pub const POSTGRES_CDC_CONNECTOR: &str = Postgres::CDC_CONNECTOR_NAME;
63pub const CITUS_CDC_CONNECTOR: &str = Citus::CDC_CONNECTOR_NAME;
64pub const MONGODB_CDC_CONNECTOR: &str = Mongodb::CDC_CONNECTOR_NAME;
65pub const SQL_SERVER_CDC_CONNECTOR: &str = SqlServer::CDC_CONNECTOR_NAME;
66
67/// Build a unique CDC table identifier from a source ID and external table name
68pub fn build_cdc_table_id(source_id: SourceId, external_table_name: &str) -> String {
69    format!("{}.{}", source_id, external_table_name)
70}
71
72pub fn normalize_simple_postgres_quoted_table_name(table_name: &str) -> Option<String> {
73    let (schema_name, table_name) = table_name.split_once('.')?;
74    let table_name = table_name.strip_prefix('"')?.strip_suffix('"')?;
75    if schema_name.is_empty()
76        || table_name.is_empty()
77        || table_name.contains('.')
78        || table_name.contains('"')
79        || table_name.contains('\\')
80        || schema_name.contains('"')
81    {
82        return None;
83    }
84
85    Some(format!("{schema_name}.{table_name}"))
86}
87
88pub trait CdcSourceTypeTrait: Send + Sync + Clone + std::fmt::Debug + 'static {
89    const CDC_CONNECTOR_NAME: &'static str;
90    fn source_type() -> CdcSourceType;
91}
92
93for_all_classified_sources!(impl_cdc_source_type);
94
95impl<'a> From<&'a str> for CdcSourceType {
96    fn from(name: &'a str) -> Self {
97        match name {
98            MYSQL_CDC_CONNECTOR => CdcSourceType::Mysql,
99            POSTGRES_CDC_CONNECTOR => CdcSourceType::Postgres,
100            CITUS_CDC_CONNECTOR => CdcSourceType::Citus,
101            MONGODB_CDC_CONNECTOR => CdcSourceType::Mongodb,
102            SQL_SERVER_CDC_CONNECTOR => CdcSourceType::SqlServer,
103            _ => CdcSourceType::Unspecified,
104        }
105    }
106}
107
108impl CdcSourceType {
109    pub fn as_str_name(&self) -> &str {
110        match self {
111            CdcSourceType::Mysql => "MySQL",
112            CdcSourceType::Postgres => "Postgres",
113            CdcSourceType::Citus => "Citus",
114            CdcSourceType::Mongodb => "MongoDB",
115            CdcSourceType::SqlServer => "SQL Server",
116            CdcSourceType::Unspecified => "Unspecified",
117        }
118    }
119}
120
121#[derive(Clone, Debug, Default)]
122pub struct CdcProperties<T: CdcSourceTypeTrait> {
123    /// Properties specified in the WITH clause by user
124    pub properties: BTreeMap<String, String>,
125
126    /// Schema of the source specified by users
127    pub table_schema: TableSchema,
128
129    /// Whether it is created by a cdc source job
130    pub is_cdc_source_job: bool,
131
132    /// For validation purpose, mark if the table is a backfill cdc table
133    pub is_backfill_table: bool,
134
135    pub _phantom: PhantomData<T>,
136}
137
138pub fn table_schema_exclude_additional_columns(table_schema: &TableSchema) -> TableSchema {
139    TableSchema {
140        columns: table_schema
141            .columns
142            .iter()
143            .filter(|col| {
144                col.additional_column
145                    .as_ref()
146                    .is_some_and(|val| val.column_type.is_none())
147            })
148            .cloned()
149            .collect(),
150        pk_indices: table_schema.pk_indices.clone(),
151    }
152}
153
154impl<T: CdcSourceTypeTrait> TryFromBTreeMap for CdcProperties<T> {
155    fn try_from_btreemap(
156        properties: BTreeMap<String, String>,
157        _deny_unknown_fields: bool,
158    ) -> ConnectorResult<Self> {
159        let is_share_source: bool = properties
160            .get(CDC_SHARING_MODE_KEY)
161            .is_some_and(|v| v == "true");
162        Ok(CdcProperties {
163            properties,
164            table_schema: Default::default(),
165            // TODO(siyuan): use serde to deserialize input hashmap
166            is_cdc_source_job: is_share_source,
167            is_backfill_table: false,
168            _phantom: PhantomData,
169        })
170    }
171}
172
173impl<T: CdcSourceTypeTrait> EnforceSecret for CdcProperties<T> {} // todo: enforce jdbc like properties
174
175impl<T: CdcSourceTypeTrait> SourceProperties for CdcProperties<T>
176where
177    DebeziumCdcSplit<T>: TryFrom<SplitImpl, Error = crate::error::ConnectorError> + Into<SplitImpl>,
178    DebeziumSplitEnumerator<T>: ListCdcSplits<CdcSourceType = T> + enumerator::CdcMonitor,
179{
180    type Split = DebeziumCdcSplit<T>;
181    type SplitEnumerator = DebeziumSplitEnumerator<T>;
182    type SplitReader = CdcSplitReader<T>;
183
184    const SOURCE_NAME: &'static str = T::CDC_CONNECTOR_NAME;
185
186    fn init_from_pb_source(&mut self, source: &PbSource) {
187        let pk_indices = source
188            .pk_column_ids
189            .iter()
190            .map(|&id| {
191                source
192                    .columns
193                    .iter()
194                    .position(|col| col.column_desc.as_ref().unwrap().column_id == id)
195                    .unwrap() as u32
196            })
197            .collect_vec();
198
199        let table_schema = PbTableSchema {
200            columns: source
201                .columns
202                .iter()
203                .flat_map(|col| &col.column_desc)
204                .filter(|col| {
205                    !matches!(
206                        col.generated_or_default_column,
207                        Some(GeneratedOrDefaultColumn::GeneratedColumn(_))
208                    )
209                })
210                .cloned()
211                .collect(),
212            pk_indices,
213        };
214        self.table_schema = table_schema;
215        if let Some(info) = source.info.as_ref() {
216            self.is_cdc_source_job = info.is_shared();
217        }
218    }
219
220    fn init_from_pb_cdc_table_desc(&mut self, table_desc: &ExternalTableDesc) {
221        let table_schema = TableSchema {
222            columns: table_desc
223                .columns
224                .iter()
225                .filter(|col| {
226                    !matches!(
227                        col.generated_or_default_column,
228                        Some(GeneratedOrDefaultColumn::GeneratedColumn(_))
229                    )
230                })
231                .cloned()
232                .collect(),
233            pk_indices: table_desc.stream_key.clone(),
234        };
235
236        self.table_schema = table_schema;
237        self.is_cdc_source_job = false;
238        self.is_backfill_table = true;
239    }
240}
241
242impl<T: CdcSourceTypeTrait> crate::source::UnknownFields for CdcProperties<T> {
243    fn unknown_fields(&self) -> HashMap<String, String> {
244        // FIXME: CDC does not handle unknown fields yet
245        HashMap::new()
246    }
247}
248
249impl<T: CdcSourceTypeTrait> CdcProperties<T> {
250    pub fn get_source_type_pb(&self) -> SourceType {
251        SourceType::from(T::source_type())
252    }
253}
254
255pub const INVALID_CDC_SPLIT_ASSIGNMENT_GENERATION_ID: u64 = 0;
256pub const INITIAL_CDC_SPLIT_ASSIGNMENT_GENERATION_ID: u64 = 1;
257
258#[derive(Clone, Debug, PartialEq, Default)]
259pub struct CdcTableSnapshotSplitAssignmentWithGeneration {
260    pub splits: HashMap<ActorId, (Vec<CdcTableSnapshotSplitRaw>, u64)>,
261}
262
263impl CdcTableSnapshotSplitAssignmentWithGeneration {
264    pub fn new(splits: HashMap<ActorId, (Vec<CdcTableSnapshotSplitRaw>, u64)>) -> Self {
265        Self { splits }
266    }
267
268    pub fn empty() -> Self {
269        Self {
270            splits: HashMap::default(),
271        }
272    }
273}
274
275pub fn build_cdc_table_snapshot_split(s: &CdcTableSnapshotSplitRaw) -> PbCdcTableSnapshotSplit {
276    PbCdcTableSnapshotSplit {
277        split_id: s.split_id,
278        left_bound_inclusive: s.left_bound_inclusive.clone(),
279        right_bound_exclusive: s.right_bound_exclusive.clone(),
280    }
281}
282
283pub fn build_actor_cdc_table_snapshot_splits_with_generation(
284    pb_cdc_table_snapshot_split_assignment: PbCdcTableSnapshotSplitsWithGeneration,
285) -> CdcTableSnapshotSplitAssignmentWithGeneration {
286    let splits = pb_cdc_table_snapshot_split_assignment
287        .splits
288        .into_iter()
289        .map(|(actor_id, splits)| {
290            let generation = splits.generation;
291            let splits = splits
292                .splits
293                .into_iter()
294                .map(|s| CdcTableSnapshotSplitRaw {
295                    split_id: s.split_id,
296                    left_bound_inclusive: s.left_bound_inclusive,
297                    right_bound_exclusive: s.right_bound_exclusive,
298                })
299                .collect();
300            (actor_id, (splits, generation))
301        })
302        .collect();
303    CdcTableSnapshotSplitAssignmentWithGeneration { splits }
304}
305
306#[derive(Debug, Clone, Hash, PartialEq)]
307pub struct CdcScanOptions {
308    /// Used by Used in non-parallel backfill, i.e. backfill V1.
309    pub disable_backfill: bool,
310    /// Used by non-parallelized backfill. The frequency of snapshot read resets for consuming the WAL backlog.
311    pub snapshot_barrier_interval: u32,
312    /// Used by non-parallelized backfill. The number of rows to fetch in a single batch when reading from an external table.
313    pub snapshot_batch_size: u32,
314    /// Used by parallelized backfill, i.e. backfill V2. The initial parallelism of parallel backfill.
315    pub backfill_parallelism: u32,
316    /// Used by parallelized backfill. The estimated number of rows per split used in splits generation.
317    pub backfill_num_rows_per_split: u64,
318    /// Used by parallelized backfill. For supported split column data type, assume an uniform distribution and adopt a much faster splits generation method.
319    pub backfill_as_even_splits: bool,
320    /// Used by parallelized backfill. Specify the index of primary key column to use as split column.
321    pub backfill_split_pk_column_index: u32,
322}
323
324impl Default for CdcScanOptions {
325    fn default() -> Self {
326        Self {
327            disable_backfill: false,
328            snapshot_barrier_interval: 10,
329            snapshot_batch_size: 1000,
330            // 0 means disable backfill v2.
331            backfill_parallelism: 0,
332            backfill_num_rows_per_split: 100_000,
333            backfill_as_even_splits: true,
334            backfill_split_pk_column_index: 0,
335        }
336    }
337}
338
339impl CdcScanOptions {
340    pub fn to_proto(&self) -> StreamCdcScanOptions {
341        StreamCdcScanOptions {
342            disable_backfill: self.disable_backfill,
343            snapshot_barrier_interval: self.snapshot_barrier_interval,
344            snapshot_batch_size: self.snapshot_batch_size,
345            backfill_parallelism: self.backfill_parallelism,
346            backfill_num_rows_per_split: self.backfill_num_rows_per_split,
347            backfill_as_even_splits: self.backfill_as_even_splits,
348            backfill_split_pk_column_index: self.backfill_split_pk_column_index,
349        }
350    }
351
352    pub fn from_proto(proto: &StreamCdcScanOptions) -> Self {
353        Self {
354            disable_backfill: proto.disable_backfill,
355            snapshot_barrier_interval: proto.snapshot_barrier_interval,
356            snapshot_batch_size: proto.snapshot_batch_size,
357            backfill_parallelism: proto.backfill_parallelism,
358            backfill_num_rows_per_split: proto.backfill_num_rows_per_split,
359            backfill_as_even_splits: proto.backfill_as_even_splits,
360            backfill_split_pk_column_index: proto.backfill_split_pk_column_index,
361        }
362    }
363
364    pub fn is_parallelized_backfill(&self) -> bool {
365        !self.disable_backfill
366            && self.backfill_num_rows_per_split > 0
367            && self.backfill_parallelism > 0
368    }
369}
370
371#[cfg(test)]
372mod tests {
373    use super::*;
374
375    #[test]
376    fn test_normalize_simple_postgres_quoted_table_name() {
377        assert_eq!(
378            normalize_simple_postgres_quoted_table_name(r#"public."TableName""#).as_deref(),
379            Some("public.TableName")
380        );
381        assert_eq!(
382            normalize_simple_postgres_quoted_table_name(r#"public."pg_my_table""#).as_deref(),
383            Some("public.pg_my_table")
384        );
385        assert_eq!(
386            normalize_simple_postgres_quoted_table_name("public.table"),
387            None
388        );
389        assert_eq!(
390            normalize_simple_postgres_quoted_table_name(r#""Mixed.Schema"."TableName""#),
391            None
392        );
393        assert_eq!(
394            normalize_simple_postgres_quoted_table_name(r#"public."table.name""#),
395            None
396        );
397        assert_eq!(
398            normalize_simple_postgres_quoted_table_name(r#"public."table\name""#),
399            None
400        );
401    }
402}
risingwave_connector/source/cdc/mod.rs

risingwave_connector/source/cdc/
mod.rs
