CI found this [1]
[ 382620 ] {} <Error> Application: Caught exception while loading metadata: Code: 60. DB::Exception: Table test_25.join doesn\'t exist: While processing _CAST(joinGet(test_25.join, \'m\', CAST(\'42\', \'int\')) AS m_tmp_alter1351896088559986425, \'Int32\') AS m: default expression and column type are incomp
[ 382620 ] {} <Error> Application: Code: 60. DB::Exception: Table test_25.join doesn\'t exist: While processing _CAST(joinGet(test_25.join, \'m\', CAST(\'42\', \'int\')) AS m_tmp_alter1351896088559986425, \'Int32\') AS m: default expression and column type are incompatible.: Cannot attach table `test_25`.`t
[1]: https://s3.amazonaws.com/clickhouse-test-reports/48670/d7f865037266ed87538cf4df7ec7e8165681871b/stress_test__asan_.html
The problem here is that after failed RENAME dependencies got lost:
[4e937f39d866] 2023.04.12 00:51:06.833624 [ 13419 ] {61429225-64cb-4fce-b60d-01e0dac6e52c} <Error> executeQuery: Code: 241. DB::Exception: Memory limit (total) exceeded: would use 34.65 GiB (attempt to allocate chunk of 2097419 bytes), maximum: 34.29 GiB. OvercommitTracker decision: Memory overcommit has freed not enough memory. (MEMORY_LIMIT_EXCEEDED) (version 23.4.1.1) (from [::1]:45710) (comment: 01160_table_dependencies.sh) (in query: create database test_25_1), Stack trace (when copying this message, always include the lines below):
[4e937f39d866] 2023.04.12 00:51:07.351914 [ 5151 ] {66d8bdd4-668e-4239-a8af-6b8f17bb5222} <Error> executeQuery: Code: 81. DB::Exception: Database test_25_1 doesn't exist. (UNKNOWN_DATABASE) (version 23.4.1.1) (from [::1]:45762) (comment: 01160_table_dependencies.sh) (in query: rename table t to test_25_1.t), Stack trace (when copying this message, always include the lines below):
And from the test output:
OK
OK
OK
OK
a [] [] []
data_02344 [] [] []
date_table [] [] []
dict1 [] ['dict_src'] ['join']
dict2 [] ['join'] []
dict_src [] [] ['dict1']
dist_02346 [] [] []
join [] ['dict1'] ['dict2','s']
^^^^^^^^^^^^^ no "t"
Signed-off-by: Azat Khuzhin <a.khuzhin@semrush.com>
* Rename DDLDependencyVisitor -> DDLLoadingDependencyVisitor.
* Move building a loading graph to TablesLoader.
* Implement referential dependencies for tables and use them
when restoring tables from a backup.
* Remove StorageID::operator < (because of its inconsistency with ==).
* Add new tests.
* Fix test.
* Fix memory leak.
Co-authored-by: Nikita Mikhaylov <mikhaylovnikitka@gmail.com>
In this version of the databaseReplicated sequential persistent zk nodes
are used to order DDL queries. Db replicated ddl queries are executed
in the backgrould pool no matter whether it's proposed by the same
replica or not.
