| 1 | #include <iomanip> | 
|---|
| 2 |  | 
|---|
| 3 | #include <Core/Settings.h> | 
|---|
| 4 | #include <Databases/DatabaseOnDisk.h> | 
|---|
| 5 | #include <Databases/DatabaseOrdinary.h> | 
|---|
| 6 | #include <Databases/DatabasesCommon.h> | 
|---|
| 7 | #include <IO/ReadBufferFromFile.h> | 
|---|
| 8 | #include <IO/ReadHelpers.h> | 
|---|
| 9 | #include <IO/WriteBufferFromFile.h> | 
|---|
| 10 | #include <IO/WriteHelpers.h> | 
|---|
| 11 | #include <Interpreters/Context.h> | 
|---|
| 12 | #include <Interpreters/InterpreterCreateQuery.h> | 
|---|
| 13 | #include <Parsers/ASTCreateQuery.h> | 
|---|
| 14 | #include <Parsers/ParserCreateQuery.h> | 
|---|
| 15 | #include <Storages/StorageFactory.h> | 
|---|
| 16 | #include <Parsers/parseQuery.h> | 
|---|
| 17 | #include <Parsers/formatAST.h> | 
|---|
| 18 | #include <Parsers/ASTSetQuery.h> | 
|---|
| 19 | #include <TableFunctions/TableFunctionFactory.h> | 
|---|
| 20 |  | 
|---|
| 21 | #include <Parsers/queryToString.h> | 
|---|
| 22 |  | 
|---|
| 23 | #include <Poco/DirectoryIterator.h> | 
|---|
| 24 | #include <Poco/Event.h> | 
|---|
| 25 | #include <Common/Stopwatch.h> | 
|---|
| 26 | #include <Common/quoteString.h> | 
|---|
| 27 | #include <Common/ThreadPool.h> | 
|---|
| 28 | #include <Common/escapeForFileName.h> | 
|---|
| 29 | #include <Common/typeid_cast.h> | 
|---|
| 30 | #include <common/logger_useful.h> | 
|---|
| 31 | #include <ext/scope_guard.h> | 
|---|
| 32 |  | 
|---|
| 33 |  | 
|---|
| 34 | namespace DB | 
|---|
| 35 | { | 
|---|
| 36 |  | 
|---|
| 37 | namespace ErrorCodes | 
|---|
| 38 | { | 
|---|
| 39 | extern const int CANNOT_CREATE_TABLE_FROM_METADATA; | 
|---|
| 40 | extern const int CANNOT_CREATE_DICTIONARY_FROM_METADATA; | 
|---|
| 41 | extern const int EMPTY_LIST_OF_COLUMNS_PASSED; | 
|---|
| 42 | extern const int CANNOT_PARSE_TEXT; | 
|---|
| 43 | } | 
|---|
| 44 |  | 
|---|
| 45 |  | 
|---|
| 46 | static constexpr size_t PRINT_MESSAGE_EACH_N_OBJECTS = 256; | 
|---|
| 47 | static constexpr size_t PRINT_MESSAGE_EACH_N_SECONDS = 5; | 
|---|
| 48 | static constexpr size_t METADATA_FILE_BUFFER_SIZE = 32768; | 
|---|
| 49 |  | 
|---|
| 50 |  | 
|---|
| 51 | namespace | 
|---|
| 52 | { | 
|---|
| 53 | void tryAttachTable( | 
|---|
| 54 | Context & context, | 
|---|
| 55 | const ASTCreateQuery & query, | 
|---|
| 56 | DatabaseOrdinary & database, | 
|---|
| 57 | const String & database_name, | 
|---|
| 58 | bool has_force_restore_data_flag) | 
|---|
| 59 | { | 
|---|
| 60 | assert(!query.is_dictionary); | 
|---|
| 61 | try | 
|---|
| 62 | { | 
|---|
| 63 | String table_name; | 
|---|
| 64 | StoragePtr table; | 
|---|
| 65 | std::tie(table_name, table) | 
|---|
| 66 | = createTableFromAST(query, database_name, database.getTableDataPath(query), context, has_force_restore_data_flag); | 
|---|
| 67 | database.attachTable(table_name, table); | 
|---|
| 68 | } | 
|---|
| 69 | catch (const Exception & e) | 
|---|
| 70 | { | 
|---|
| 71 | throw Exception( | 
|---|
| 72 | "Cannot attach table '"+ query.table + "' from query "+ serializeAST(query) | 
|---|
| 73 | + ". Error: "+ DB::getCurrentExceptionMessage(true), | 
|---|
| 74 | e, | 
|---|
| 75 | DB::ErrorCodes::CANNOT_CREATE_TABLE_FROM_METADATA); | 
|---|
| 76 | } | 
|---|
| 77 | } | 
|---|
| 78 |  | 
|---|
| 79 |  | 
|---|
| 80 | void tryAttachDictionary( | 
|---|
| 81 | Context & context, | 
|---|
| 82 | const ASTCreateQuery & query, | 
|---|
| 83 | DatabaseOrdinary & database) | 
|---|
| 84 | { | 
|---|
| 85 | assert(query.is_dictionary); | 
|---|
| 86 | try | 
|---|
| 87 | { | 
|---|
| 88 | database.attachDictionary(query.table, context); | 
|---|
| 89 | } | 
|---|
| 90 | catch (const Exception & e) | 
|---|
| 91 | { | 
|---|
| 92 | throw Exception( | 
|---|
| 93 | "Cannot create dictionary '"+ query.table + "' from query "+ serializeAST(query) | 
|---|
| 94 | + ". Error: "+ DB::getCurrentExceptionMessage(true), | 
|---|
| 95 | e, | 
|---|
| 96 | DB::ErrorCodes::CANNOT_CREATE_DICTIONARY_FROM_METADATA); | 
|---|
| 97 | } | 
|---|
| 98 | } | 
|---|
| 99 |  | 
|---|
| 100 |  | 
|---|
| 101 | void logAboutProgress(Poco::Logger * log, size_t processed, size_t total, AtomicStopwatch & watch) | 
|---|
| 102 | { | 
|---|
| 103 | if (processed % PRINT_MESSAGE_EACH_N_OBJECTS == 0 || watch.compareAndRestart(PRINT_MESSAGE_EACH_N_SECONDS)) | 
|---|
| 104 | { | 
|---|
| 105 | LOG_INFO(log, std::fixed << std::setprecision(2) << processed * 100.0 / total << "%"); | 
|---|
| 106 | watch.restart(); | 
|---|
| 107 | } | 
|---|
| 108 | } | 
|---|
| 109 | } | 
|---|
| 110 |  | 
|---|
| 111 |  | 
|---|
| 112 | DatabaseOrdinary::DatabaseOrdinary(const String & name_, const String & metadata_path_, const Context & context_) | 
|---|
| 113 | : DatabaseWithDictionaries(name_, metadata_path_, "DatabaseOrdinary ("+ name_ + ")") | 
|---|
| 114 | { | 
|---|
| 115 | Poco::File(context_.getPath() + getDataPath()).createDirectories(); | 
|---|
| 116 | } | 
|---|
| 117 |  | 
|---|
| 118 |  | 
|---|
| 119 | void DatabaseOrdinary::loadStoredObjects( | 
|---|
| 120 | Context & context, | 
|---|
| 121 | bool has_force_restore_data_flag) | 
|---|
| 122 | { | 
|---|
| 123 |  | 
|---|
| 124 | /** Tables load faster if they are loaded in sorted (by name) order. | 
|---|
| 125 | * Otherwise (for the ext4 filesystem), `DirectoryIterator` iterates through them in some order, | 
|---|
| 126 | *  which does not correspond to order tables creation and does not correspond to order of their location on disk. | 
|---|
| 127 | */ | 
|---|
| 128 | using FileNames = std::map<std::string, ASTPtr>; | 
|---|
| 129 | FileNames file_names; | 
|---|
| 130 |  | 
|---|
| 131 | size_t total_dictionaries = 0; | 
|---|
| 132 | iterateMetadataFiles(context, [&file_names, &total_dictionaries, this](const String & file_name) | 
|---|
| 133 | { | 
|---|
| 134 | String full_path = getMetadataPath() + file_name; | 
|---|
| 135 | try | 
|---|
| 136 | { | 
|---|
| 137 | auto ast = parseQueryFromMetadata(full_path, /*throw_on_error*/ true, /*remove_empty*/false); | 
|---|
| 138 | if (ast) | 
|---|
| 139 | { | 
|---|
| 140 | auto * create_query = ast->as<ASTCreateQuery>(); | 
|---|
| 141 | file_names[file_name] = ast; | 
|---|
| 142 | total_dictionaries += create_query->is_dictionary; | 
|---|
| 143 | } | 
|---|
| 144 | } | 
|---|
| 145 | catch (const Exception & e) | 
|---|
| 146 | { | 
|---|
| 147 | throw Exception( | 
|---|
| 148 | "Cannot parse definition from metadata file "+ full_path + ". Error: "+ DB::getCurrentExceptionMessage(true), e, ErrorCodes::CANNOT_PARSE_TEXT); | 
|---|
| 149 | } | 
|---|
| 150 |  | 
|---|
| 151 | }); | 
|---|
| 152 |  | 
|---|
| 153 | size_t total_tables = file_names.size() - total_dictionaries; | 
|---|
| 154 |  | 
|---|
| 155 | LOG_INFO(log, "Total "<< total_tables << " tables and "<< total_dictionaries << " dictionaries."); | 
|---|
| 156 |  | 
|---|
| 157 | AtomicStopwatch watch; | 
|---|
| 158 | std::atomic<size_t> tables_processed{0}; | 
|---|
| 159 | std::atomic<size_t> dictionaries_processed{0}; | 
|---|
| 160 |  | 
|---|
| 161 | ThreadPool pool(SettingMaxThreads().getAutoValue()); | 
|---|
| 162 |  | 
|---|
| 163 | /// Attach tables. | 
|---|
| 164 | for (const auto & name_with_query : file_names) | 
|---|
| 165 | { | 
|---|
| 166 | const auto & create_query = name_with_query.second->as<const ASTCreateQuery &>(); | 
|---|
| 167 | if (!create_query.is_dictionary) | 
|---|
| 168 | pool.scheduleOrThrowOnError([&]() | 
|---|
| 169 | { | 
|---|
| 170 | tryAttachTable(context, create_query, *this, getDatabaseName(), has_force_restore_data_flag); | 
|---|
| 171 |  | 
|---|
| 172 | /// Messages, so that it's not boring to wait for the server to load for a long time. | 
|---|
| 173 | logAboutProgress(log, ++tables_processed, total_tables, watch); | 
|---|
| 174 | }); | 
|---|
| 175 | } | 
|---|
| 176 |  | 
|---|
| 177 | pool.wait(); | 
|---|
| 178 |  | 
|---|
| 179 | /// After all tables was basically initialized, startup them. | 
|---|
| 180 | startupTables(pool); | 
|---|
| 181 |  | 
|---|
| 182 | /// Attach dictionaries. | 
|---|
| 183 | attachToExternalDictionariesLoader(context); | 
|---|
| 184 | for (const auto & name_with_query : file_names) | 
|---|
| 185 | { | 
|---|
| 186 | auto create_query = name_with_query.second->as<const ASTCreateQuery &>(); | 
|---|
| 187 | if (create_query.is_dictionary) | 
|---|
| 188 | { | 
|---|
| 189 | tryAttachDictionary(context, create_query, *this); | 
|---|
| 190 |  | 
|---|
| 191 | /// Messages, so that it's not boring to wait for the server to load for a long time. | 
|---|
| 192 | logAboutProgress(log, ++dictionaries_processed, total_dictionaries, watch); | 
|---|
| 193 | } | 
|---|
| 194 | } | 
|---|
| 195 | } | 
|---|
| 196 |  | 
|---|
| 197 |  | 
|---|
| 198 | void DatabaseOrdinary::startupTables(ThreadPool & thread_pool) | 
|---|
| 199 | { | 
|---|
| 200 | LOG_INFO(log, "Starting up tables."); | 
|---|
| 201 |  | 
|---|
| 202 | const size_t total_tables = tables.size(); | 
|---|
| 203 | if (!total_tables) | 
|---|
| 204 | return; | 
|---|
| 205 |  | 
|---|
| 206 | AtomicStopwatch watch; | 
|---|
| 207 | std::atomic<size_t> tables_processed{0}; | 
|---|
| 208 |  | 
|---|
| 209 | auto startupOneTable = [&](const StoragePtr & table) | 
|---|
| 210 | { | 
|---|
| 211 | table->startup(); | 
|---|
| 212 | logAboutProgress(log, ++tables_processed, total_tables, watch); | 
|---|
| 213 | }; | 
|---|
| 214 |  | 
|---|
| 215 | try | 
|---|
| 216 | { | 
|---|
| 217 | for (const auto & table : tables) | 
|---|
| 218 | thread_pool.scheduleOrThrowOnError([&]() { startupOneTable(table.second); }); | 
|---|
| 219 | } | 
|---|
| 220 | catch (...) | 
|---|
| 221 | { | 
|---|
| 222 | thread_pool.wait(); | 
|---|
| 223 | throw; | 
|---|
| 224 | } | 
|---|
| 225 | thread_pool.wait(); | 
|---|
| 226 | } | 
|---|
| 227 |  | 
|---|
| 228 | void DatabaseOrdinary::alterTable( | 
|---|
| 229 | const Context & context, | 
|---|
| 230 | const String & table_name, | 
|---|
| 231 | const StorageInMemoryMetadata & metadata) | 
|---|
| 232 | { | 
|---|
| 233 | /// Read the definition of the table and replace the necessary parts with new ones. | 
|---|
| 234 | String table_metadata_path = getObjectMetadataPath(table_name); | 
|---|
| 235 | String table_metadata_tmp_path = table_metadata_path + ".tmp"; | 
|---|
| 236 | String statement; | 
|---|
| 237 |  | 
|---|
| 238 | { | 
|---|
| 239 | char in_buf[METADATA_FILE_BUFFER_SIZE]; | 
|---|
| 240 | ReadBufferFromFile in(table_metadata_path, METADATA_FILE_BUFFER_SIZE, -1, in_buf); | 
|---|
| 241 | readStringUntilEOF(statement, in); | 
|---|
| 242 | } | 
|---|
| 243 |  | 
|---|
| 244 | ParserCreateQuery parser; | 
|---|
| 245 | ASTPtr ast = parseQuery(parser, statement.data(), statement.data() + statement.size(), "in file "+ table_metadata_path, 0); | 
|---|
| 246 |  | 
|---|
| 247 | const auto & ast_create_query = ast->as<ASTCreateQuery &>(); | 
|---|
| 248 |  | 
|---|
| 249 | ASTPtr new_columns = InterpreterCreateQuery::formatColumns(metadata.columns); | 
|---|
| 250 | ASTPtr new_indices = InterpreterCreateQuery::formatIndices(metadata.indices); | 
|---|
| 251 | ASTPtr new_constraints = InterpreterCreateQuery::formatConstraints(metadata.constraints); | 
|---|
| 252 |  | 
|---|
| 253 | ast_create_query.columns_list->replace(ast_create_query.columns_list->columns, new_columns); | 
|---|
| 254 | ast_create_query.columns_list->setOrReplace(ast_create_query.columns_list->indices, new_indices); | 
|---|
| 255 | ast_create_query.columns_list->setOrReplace(ast_create_query.columns_list->constraints, new_constraints); | 
|---|
| 256 |  | 
|---|
| 257 | ASTStorage & storage_ast = *ast_create_query.storage; | 
|---|
| 258 | /// ORDER BY may change, but cannot appear, it's required construction | 
|---|
| 259 | if (metadata.order_by_ast && storage_ast.order_by) | 
|---|
| 260 | storage_ast.set(storage_ast.order_by, metadata.order_by_ast); | 
|---|
| 261 |  | 
|---|
| 262 | if (metadata.primary_key_ast) | 
|---|
| 263 | storage_ast.set(storage_ast.primary_key, metadata.primary_key_ast); | 
|---|
| 264 |  | 
|---|
| 265 | if (metadata.ttl_for_table_ast) | 
|---|
| 266 | storage_ast.set(storage_ast.ttl_table, metadata.ttl_for_table_ast); | 
|---|
| 267 |  | 
|---|
| 268 | if (metadata.settings_ast) | 
|---|
| 269 | storage_ast.set(storage_ast.settings, metadata.settings_ast); | 
|---|
| 270 |  | 
|---|
| 271 |  | 
|---|
| 272 | statement = getObjectDefinitionFromCreateQuery(ast); | 
|---|
| 273 | { | 
|---|
| 274 | WriteBufferFromFile out(table_metadata_tmp_path, statement.size(), O_WRONLY | O_CREAT | O_EXCL); | 
|---|
| 275 | writeString(statement, out); | 
|---|
| 276 | out.next(); | 
|---|
| 277 | if (context.getSettingsRef().fsync_metadata) | 
|---|
| 278 | out.sync(); | 
|---|
| 279 | out.close(); | 
|---|
| 280 | } | 
|---|
| 281 |  | 
|---|
| 282 | try | 
|---|
| 283 | { | 
|---|
| 284 | /// rename atomically replaces the old file with the new one. | 
|---|
| 285 | Poco::File(table_metadata_tmp_path).renameTo(table_metadata_path); | 
|---|
| 286 | } | 
|---|
| 287 | catch (...) | 
|---|
| 288 | { | 
|---|
| 289 | Poco::File(table_metadata_tmp_path).remove(); | 
|---|
| 290 | throw; | 
|---|
| 291 | } | 
|---|
| 292 | } | 
|---|
| 293 |  | 
|---|
| 294 | } | 
|---|
| 295 |  | 
|---|