| Bug #88299 | csv:binlog corruption, abort, An error occurred during flush stage of the commit | ||
|---|---|---|---|
| Submitted: | 30 Oct 2017 15:39 | ||
| Reporter: | Shane Bester (Platinum Quality Contributor) | Email Updates: | |
| Status: | Verified | Impact on me: | |
| Category: | MySQL Server: CSV | Severity: | S1 (Critical) |
| Version: | 5.7.17 | OS: | Windows |
| Assigned to: | CPU Architecture: | Any | |
[31 Oct 2017 5:47]
MySQL Verification Team
it looks windows specific so far...
[31 Oct 2017 7:02]
MySQL Verification Team
Since this is a windows problem, I enabled Application Verifier (appverif.exe). Here's the first report. PRIMARY_PROBLEM_CLASS: STATUS_INVALID_HANDLE BUGCHECK_STR: APPLICATION_FAULT_STATUS_INVALID_HANDLE Call Site ntdll!KiRaiseUserExceptionDispatcher+0x3a vfbasics+0x19558 KERNELBASE!SetFilePointerEx+0x93 kernel32!SetFilePointerExStub+0x9 mysqld!my_win_lseek+0x29 [.\mysql-5.7.20\mysys\my_winfile.c @ 409] mysqld!my_seek+0x26 [.\mysql-5.7.20\mysys\my_seek.c @ 64] mysqld!inline_mysql_file_seek+0xaf [.\mysql-5.7.20\include\mysql\psi\mysql_file.h @ 1243] mysqld!Transparent_file::init_buff+0x32 [.\mysql-5.7.20\storage\csv\transparent_file.cc @ 39] mysqld!ha_tina::init_data_file+0xab [.\mysql-5.7.20\storage\csv\ha_tina.cc @ 1159] mysqld!ha_tina::rnd_init+0x1e [.\mysql-5.7.20\storage\csv\ha_tina.cc @ 1199] mysqld!handler::ha_rnd_init+0x12 [.\mysql-5.7.20\sql\handler.cc @ 2897] mysqld!init_read_record+0x2d1 [.\mysql-5.7.20\sql\records.cc @ 314] mysqld!Sql_cmd_delete::mysql_delete+0xa66 [.\mysql-5.7.20\sql\sql_delete.cc @ 412] mysqld!Sql_cmd_delete::execute+0xc9 [.\mysql-5.7.20\sql\sql_delete.cc @ 1382] mysqld!mysql_execute_command+0x1865 [.\mysql-5.7.20\sql\sql_parse.cc @ 3571] mysqld!mysql_parse+0x273 [.\mysql-5.7.20\sql\sql_parse.cc @ 5583] mysqld!dispatch_command+0x713 [.\mysql-5.7.20\sql\sql_parse.cc @ 1464] mysqld!do_command+0x1ea [.\mysql-5.7.20\sql\sql_parse.cc @ 1001] mysqld!handle_connection+0x12c [.\mysql-5.7.20\sql\conn_handler\connection_handler_per_thread.cc @ 300] mysqld!pfs_spawn_thread+0x132 [.\mysql-5.7.20\storage\perfschema\pfs.cc @ 2193] mysqld!win_thread_start+0x1c [.\mysql-5.7.20\mysys\my_thread.c @ 38]
Description: Affects release builds of 5.7. Symptoms are: 1. Some user queries error out with strange errors: Error writing file 'UNOPENED' (Errcode: 9 - Bad file descriptor) 2. Inexplicable crashes like this: [ERROR] mysqld: Binary logging not possible. Message: An error occurred during flush stage of the commit. 'binlog_error_action' is set to 'ABORT_SERVER'. Hence aborting the server. 15:26:28 UTC - mysqld got exception 0x80000003 ; mysqld.exe!my_sigabrt_handler()[my_thr_init.c:449] ucrtbase.DLL!raise() ucrtbase.DLL!abort() mysqld.exe!exec_binlog_error_action_abort()[binlog.cc:1858] mysqld.exe!MYSQL_BIN_LOG::handle_binlog_flush_or_sync_error()[binlog.cc:9105] mysqld.exe!MYSQL_BIN_LOG::ordered_commit()[binlog.cc:9325] mysqld.exe!MYSQL_BIN_LOG::commit()[binlog.cc:8529] mysqld.exe!ha_commit_trans()[handler.cc:1796] mysqld.exe!trans_commit_stmt()[transaction.cc:459] mysqld.exe!mysql_execute_command()[sql_parse.cc:5005] mysqld.exe!mysql_parse()[sql_parse.cc:5588] mysqld.exe!dispatch_command()[sql_parse.cc:1460] mysqld.exe!do_command()[sql_parse.cc:1000] mysqld.exe!handle_connection()[connection_handler_per_thread.cc:300] mysqld.exe!pfs_spawn_thread()[pfs.cc:2193] mysqld.exe!win_thread_start()[my_thread.c:38] 3. When trying to restart mysqld, it may fail to recognize its own binary logs, or the binlog .index file is containing trash... [Note] Plugin 'FEDERATED' is disabled. [ERROR] Binlog has bad magic number; It's not a binary log file that can be used by this version of MySQL [ERROR] Can't init tc log [ERROR] Aborting How to repeat: Testcase must be run like this, exactly. 1. mysqld --max-binlog-size=1M --binlog-format=STATEMENT --log-bin --server-id=1 2. Create the schema: reset master; drop database if exists test; create database test; use test; drop table if exists `t1`; create table `t1` (`c0000` varbinary(100) not null) engine=csv; create table t2 like t1; 3. Import the testcase: mysql -uroot --force test --binary-mode < /tmp/t.sql Repeat as needed, since sometimes it doesn't crash.. If above does not satisfactorily repeat problems, simply run a random DML testcase on the table with random data and insert/delete/update/replace/select.