be/src/storage/schema_change/schema_change.h
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #pragma once |
19 | | |
20 | | #include <butil/macros.h> |
21 | | #include <fmt/format.h> |
22 | | #include <gen_cpp/Descriptors_types.h> |
23 | | #include <glog/logging.h> |
24 | | #include <stddef.h> |
25 | | #include <stdint.h> |
26 | | |
27 | | #include <memory> |
28 | | #include <ostream> |
29 | | #include <set> |
30 | | #include <shared_mutex> |
31 | | #include <string> |
32 | | #include <unordered_map> |
33 | | #include <unordered_set> |
34 | | #include <utility> |
35 | | #include <vector> |
36 | | |
37 | | #include "common/config.h" |
38 | | #include "common/object_pool.h" |
39 | | #include "common/status.h" |
40 | | #include "core/data_type/data_type.h" |
41 | | #include "runtime/descriptors.h" |
42 | | #include "runtime/memory/mem_tracker.h" |
43 | | #include "runtime/runtime_state.h" |
44 | | #include "storage/olap_common.h" |
45 | | #include "storage/rowset/pending_rowset_helper.h" |
46 | | #include "storage/rowset/rowset.h" |
47 | | #include "storage/rowset/rowset_reader.h" |
48 | | #include "storage/rowset/rowset_writer.h" |
49 | | #include "storage/schema_change/column_mapping.h" |
50 | | #include "storage/storage_engine.h" |
51 | | #include "storage/tablet/tablet.h" |
52 | | #include "storage/tablet/tablet_fwd.h" |
53 | | #include "storage/tablet/tablet_schema.h" |
54 | | |
55 | | namespace doris { |
56 | | class DeleteHandler; |
57 | | class TAlterInvertedIndexReq; |
58 | | class TAlterTabletReqV2; |
59 | | class TExpr; |
60 | | enum AlterTabletType : int; |
61 | | enum RowsetTypePB : int; |
62 | | enum SegmentsOverlapPB : int; |
63 | | |
64 | | class Block; |
65 | | class OlapBlockDataConvertor; |
66 | | |
67 | | class BlockChanger { |
68 | | public: |
69 | | BlockChanger(TabletSchemaSPtr tablet_schema, DescriptorTbl desc_tbl, |
70 | | std::shared_ptr<RuntimeState> state); |
71 | | |
72 | | ~BlockChanger(); |
73 | | |
74 | | ColumnMapping* get_mutable_column_mapping(size_t column_index); |
75 | | |
76 | | Status change_block(Block* ref_block, Block* new_block) const; |
77 | | |
78 | 0 | void set_where_expr(const std::shared_ptr<TExpr>& where_expr) { _where_expr = where_expr; } |
79 | | |
80 | 1 | void set_type(AlterTabletType type) { _type = type; } |
81 | | |
82 | 1 | void set_compatible_version(int32_t version) noexcept { _fe_compatible_version = version; } |
83 | | |
84 | 0 | bool has_where() const { return _where_expr != nullptr; } |
85 | | |
86 | | private: |
87 | | static Status _check_cast_valid(ColumnPtr ref_column, ColumnPtr new_column); |
88 | | |
89 | | // @brief column-mapping specification of new schema |
90 | | SchemaMapping _schema_mapping; |
91 | | |
92 | | DescriptorTbl _desc_tbl; |
93 | | |
94 | | std::shared_ptr<TExpr> _where_expr; |
95 | | |
96 | | AlterTabletType _type; |
97 | | |
98 | | int32_t _fe_compatible_version = -1; |
99 | | |
100 | | std::shared_ptr<RuntimeState> _state; |
101 | | }; |
102 | | |
103 | | class SchemaChange { |
104 | | public: |
105 | 1 | SchemaChange() = default; |
106 | 1 | virtual ~SchemaChange() = default; |
107 | | |
108 | | virtual Status process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
109 | | BaseTabletSPtr new_tablet, BaseTabletSPtr base_tablet, |
110 | | TabletSchemaSPtr base_tablet_schema, |
111 | 0 | TabletSchemaSPtr new_tablet_schema) { |
112 | 0 | if (rowset_reader->rowset()->empty() || rowset_reader->rowset()->num_rows() == 0) { |
113 | 0 | RETURN_IF_ERROR(rowset_writer->flush()); |
114 | 0 | return Status::OK(); |
115 | 0 | } |
116 | | |
117 | 0 | _filtered_rows = 0; |
118 | 0 | _merged_rows = 0; |
119 | 0 | RETURN_IF_ERROR_OR_CATCH_EXCEPTION(_inner_process(rowset_reader, rowset_writer, new_tablet, |
120 | 0 | base_tablet_schema, new_tablet_schema)); |
121 | | |
122 | | // Check row num changes |
123 | 0 | if (!_check_row_nums(rowset_reader, *rowset_writer)) { |
124 | 0 | return Status::Error<ErrorCode::ALTER_STATUS_ERR>("SchemaChange check row nums failed"); |
125 | 0 | } |
126 | | |
127 | 0 | LOG(INFO) << "all row nums. source_rows=" << rowset_reader->rowset()->num_rows() |
128 | 0 | << ", source_filtered_rows=" << rowset_reader->filtered_rows() |
129 | 0 | << ", source_merged_rows=" << rowset_reader->merged_rows() |
130 | 0 | << ", merged_rows=" << merged_rows() << ", filtered_rows=" << filtered_rows() |
131 | 0 | << ", new_index_rows=" << rowset_writer->num_rows() |
132 | 0 | << ", writer_filtered_rows=" << rowset_writer->num_rows_filtered(); |
133 | 0 | return Status::OK(); |
134 | 0 | } |
135 | | |
136 | 0 | uint64_t filtered_rows() const { return _filtered_rows; } |
137 | | |
138 | 0 | uint64_t merged_rows() const { return _merged_rows; } |
139 | | |
140 | | protected: |
141 | 0 | void _add_filtered_rows(uint64_t filtered_rows) { _filtered_rows += filtered_rows; } |
142 | | |
143 | 0 | void _add_merged_rows(uint64_t merged_rows) { _merged_rows += merged_rows; } |
144 | | |
145 | | virtual Status _inner_process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
146 | | BaseTabletSPtr new_tablet, TabletSchemaSPtr base_tablet_schema, |
147 | 0 | TabletSchemaSPtr new_tablet_schema) { |
148 | 0 | return Status::NotSupported("inner process unsupported."); |
149 | 0 | } |
150 | | |
151 | 0 | virtual bool _check_row_nums(RowsetReaderSharedPtr reader, const RowsetWriter& writer) const { |
152 | 0 | if (reader->rowset()->num_rows() - reader->filtered_rows() - reader->merged_rows() != |
153 | 0 | writer.num_rows() + writer.num_rows_filtered() + _merged_rows + _filtered_rows) { |
154 | 0 | LOG(WARNING) << "fail to check row num! " |
155 | 0 | << "source_rows=" << reader->rowset()->num_rows() |
156 | 0 | << ", source_filtered_rows=" << reader->filtered_rows() |
157 | 0 | << ", source_merged_rows=" << reader->merged_rows() |
158 | 0 | << ", written_rows=" << writer.num_rows() |
159 | 0 | << ", writer_filtered_rows=" << writer.num_rows_filtered() |
160 | 0 | << ", merged_rows=" << merged_rows() |
161 | 0 | << ", filtered_rows=" << filtered_rows(); |
162 | 0 | if (!config::ignore_schema_change_check) { |
163 | 0 | return false; |
164 | 0 | } |
165 | 0 | } |
166 | 0 | return true; |
167 | 0 | } |
168 | | |
169 | | private: |
170 | | uint64_t _filtered_rows {}; |
171 | | uint64_t _merged_rows {}; |
172 | | |
173 | | protected: |
174 | | std::vector<bool> _row_same_bit; |
175 | | }; |
176 | | |
177 | | class LinkedSchemaChange : public SchemaChange { |
178 | | public: |
179 | 0 | LinkedSchemaChange() = default; |
180 | 0 | ~LinkedSchemaChange() override = default; |
181 | | |
182 | | Status process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
183 | | BaseTabletSPtr new_tablet, BaseTabletSPtr base_tablet, |
184 | | TabletSchemaSPtr base_tablet_schema, |
185 | | TabletSchemaSPtr new_tablet_schema) override; |
186 | | |
187 | | private: |
188 | | DISALLOW_COPY_AND_ASSIGN(LinkedSchemaChange); |
189 | | }; |
190 | | |
191 | | class VSchemaChangeDirectly : public SchemaChange { |
192 | | public: |
193 | 1 | VSchemaChangeDirectly(const BlockChanger& changer) : _changer(changer) {} |
194 | | |
195 | | private: |
196 | | Status _inner_process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
197 | | BaseTabletSPtr new_tablet, TabletSchemaSPtr base_tablet_schema, |
198 | | TabletSchemaSPtr new_tablet_schema) override; |
199 | | |
200 | 0 | bool _check_row_nums(RowsetReaderSharedPtr reader, const RowsetWriter& writer) const override { |
201 | 0 | return _changer.has_where() || SchemaChange::_check_row_nums(reader, writer); |
202 | 0 | } |
203 | | |
204 | | const BlockChanger& _changer; |
205 | | }; |
206 | | |
207 | | class VBaseSchemaChangeWithSorting : public SchemaChange { |
208 | | public: |
209 | | VBaseSchemaChangeWithSorting(const BlockChanger& changer, size_t memory_limitation); |
210 | 0 | ~VBaseSchemaChangeWithSorting() override = default; |
211 | | |
212 | | Status _inner_process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
213 | | BaseTabletSPtr new_tablet, TabletSchemaSPtr base_tablet_schema, |
214 | | TabletSchemaSPtr new_tablet_schema) override; |
215 | | |
216 | | virtual Result<RowsetSharedPtr> _internal_sorting( |
217 | | const std::vector<std::unique_ptr<Block>>& blocks, const Version& temp_delta_versions, |
218 | | int64_t newest_write_timestamp, BaseTabletSPtr new_tablet, RowsetTypePB new_rowset_type, |
219 | | SegmentsOverlapPB segments_overlap, TabletSchemaSPtr new_tablet_schema); |
220 | | |
221 | | Status _external_sorting(std::vector<RowsetSharedPtr>& src_rowsets, RowsetWriter* rowset_writer, |
222 | | BaseTabletSPtr new_tablet, TabletSchemaSPtr new_tablet_schema); |
223 | | |
224 | | protected: |
225 | | // for external sorting |
226 | | // src_rowsets to store the rowset generated by internal sorting |
227 | | std::vector<RowsetSharedPtr> _src_rowsets; |
228 | | |
229 | | private: |
230 | 0 | bool _check_row_nums(RowsetReaderSharedPtr reader, const RowsetWriter& writer) const override { |
231 | 0 | return _changer.has_where() || SchemaChange::_check_row_nums(reader, writer); |
232 | 0 | } |
233 | | |
234 | | const BlockChanger& _changer; |
235 | | size_t _memory_limitation; |
236 | | Version _temp_delta_versions; |
237 | | std::unique_ptr<MemTracker> _mem_tracker; |
238 | | }; |
239 | | |
240 | | // @breif schema change with sorting |
241 | | // Mixin for local StorageEngine |
242 | | class VLocalSchemaChangeWithSorting final : public VBaseSchemaChangeWithSorting { |
243 | | public: |
244 | | VLocalSchemaChangeWithSorting(const BlockChanger& changer, size_t memory_limitation, |
245 | | StorageEngine& local_storage_engine) |
246 | 0 | : VBaseSchemaChangeWithSorting(changer, memory_limitation), |
247 | 0 | _local_storage_engine(local_storage_engine) {} |
248 | 0 | ~VLocalSchemaChangeWithSorting() override = default; |
249 | | |
250 | | Status _inner_process(RowsetReaderSharedPtr rowset_reader, RowsetWriter* rowset_writer, |
251 | | BaseTabletSPtr new_tablet, TabletSchemaSPtr base_tablet_schema, |
252 | | TabletSchemaSPtr new_tablet_schema) override; |
253 | | |
254 | | Result<RowsetSharedPtr> _internal_sorting( |
255 | | const std::vector<std::unique_ptr<Block>>& blocks, const Version& temp_delta_versions, |
256 | | int64_t newest_write_timestamp, BaseTabletSPtr new_tablet, RowsetTypePB new_rowset_type, |
257 | | SegmentsOverlapPB segments_overlap, TabletSchemaSPtr new_tablet_schema) override; |
258 | | |
259 | | private: |
260 | | StorageEngine& _local_storage_engine; |
261 | | std::vector<PendingRowsetGuard> _pending_rs_guards; |
262 | | }; |
263 | | |
264 | | struct AlterMaterializedViewParam { |
265 | | std::string column_name; |
266 | | std::string origin_column_name; |
267 | | std::shared_ptr<TExpr> expr; |
268 | | }; |
269 | | |
270 | | struct SchemaChangeParams { |
271 | | AlterTabletType alter_tablet_type; |
272 | | bool enable_unique_key_merge_on_write = false; |
273 | | std::vector<RowsetReaderSharedPtr> ref_rowset_readers; |
274 | | DeleteHandler* delete_handler = nullptr; |
275 | | std::unordered_map<std::string, AlterMaterializedViewParam> materialized_params_map; |
276 | | DescriptorTbl* desc_tbl = nullptr; |
277 | | ObjectPool pool; |
278 | | int32_t be_exec_version; |
279 | | std::string vault_id; |
280 | | bool output_to_file_cache; |
281 | | std::shared_ptr<RuntimeState> runtime_state; |
282 | | }; |
283 | | |
284 | | class SchemaChangeJob { |
285 | | public: |
286 | | SchemaChangeJob(StorageEngine& local_storage_engine, const TAlterTabletReqV2& request, |
287 | | const std::string& job_id); |
288 | | Status process_alter_tablet(const TAlterTabletReqV2& request); |
289 | | |
290 | | bool tablet_in_converting(int64_t tablet_id); |
291 | | |
292 | | static Status parse_request(const SchemaChangeParams& sc_params, |
293 | | TabletSchema* base_tablet_schema, TabletSchema* new_tablet_schema, |
294 | | BlockChanger* changer, bool* sc_sorting, bool* sc_directly); |
295 | | |
296 | | private: |
297 | | std::unique_ptr<SchemaChange> _get_sc_procedure(const BlockChanger& changer, bool sc_sorting, |
298 | 0 | bool sc_directly, int64_t mem_limit) { |
299 | 0 | if (sc_sorting) { |
300 | 0 | return std::make_unique<VLocalSchemaChangeWithSorting>(changer, mem_limit, |
301 | 0 | _local_storage_engine); |
302 | 0 | } |
303 | | |
304 | 0 | if (sc_directly) { |
305 | 0 | return std::make_unique<VSchemaChangeDirectly>(changer); |
306 | 0 | } |
307 | | |
308 | 0 | return std::make_unique<LinkedSchemaChange>(); |
309 | 0 | } |
310 | | |
311 | | Status _get_versions_to_be_changed(std::vector<Version>* versions_to_be_changed, |
312 | | RowsetSharedPtr* max_rowset); |
313 | | |
314 | | Status _do_process_alter_tablet(const TAlterTabletReqV2& request); |
315 | | |
316 | | Status _validate_alter_result(const TAlterTabletReqV2& request); |
317 | | |
318 | | Status _convert_historical_rowsets(const SchemaChangeParams& sc_params, |
319 | | int64_t* real_alter_version); |
320 | | |
321 | | // Initialization Settings for creating a default value |
322 | | static Status _init_column_mapping(ColumnMapping* column_mapping, |
323 | | const TabletColumn& column_schema, const std::string& value); |
324 | | |
325 | | Status _calc_delete_bitmap_for_mow_table(int64_t alter_version); |
326 | | |
327 | | StorageEngine& _local_storage_engine; |
328 | | TabletSharedPtr _base_tablet; |
329 | | TabletSharedPtr _new_tablet; |
330 | | TabletSchemaSPtr _base_tablet_schema; |
331 | | TabletSchemaSPtr _new_tablet_schema; |
332 | | std::shared_mutex _mutex; |
333 | | std::unordered_set<int64_t> _tablet_ids_in_converting; |
334 | | std::set<std::string> _supported_functions; |
335 | | std::string _job_id; |
336 | | }; |
337 | | } // namespace doris |