MariaDB: ERROR 2013 (HY000): Lost connection persists only for one database - mysql

Question

I have several databases:
- curr
- curr_add
and many others on the same MariaDB instance. When I submit an sql to curr_add I get connection and the results. When I send the query to curr I get the error:
MariaDB: ERROR 2013 (HY000): Lost connection to MySQL server during query
and the service stops. When I restart the service each time I want to access the database curr the serivice stops. This is not the case for all other databases of the same instance. The instance and databases exists since one year and I had never such problems. Is there a way to find out what is the problem and how to solve it?
Working environment:
Windows 7
MariaDB 10.2.6
I entered as root: mysql -uroot -h localhost -p
Update (1):
I have access to database information_schema. For example, I can count the rows of table system_variables: SELECT COUNT(*) FROM system_variables. But if I try the same with table columns: SELECT COUNT(*) FROM columns, the connection will be lost (see figure).
I submit:
SELECT COUNT(*) FROM tables
or
SELECT table_schema, table_name FROM tables
I get results.
But if I submit
SELECT * FROM tables
or
SELECT table_schema, table_name, engine, table_rows FROM tables LIMIT 10
the connection will be lost.
Summary: some database disconnect the mysql-server and also some columns of tables.
Update (2): error informations from file .err
The manual page at http://dev.mysql.com/doc/mysql/en/crashing.html contains
information that should help you find out what is causing the crash.
2017-11-28 19:42:43 7820 [Note] InnoDB: Mutexes and rw_locks use Windows interlocked functions
2017-11-28 19:42:43 7820 [Note] InnoDB: Uses event mutexes
2017-11-28 19:42:43 7820 [Note] InnoDB: Compressed tables use zlib 1.2.3
2017-11-28 19:42:43 7820 [Note] InnoDB: Number of pools: 1
2017-11-28 19:42:43 7820 [Note] InnoDB: Using generic crc32 instructions
2017-11-28 19:42:43 7820 [Note] InnoDB: Initializing buffer pool, total size = 14G, instances = 8, chunk size = 128M
2017-11-28 19:42:44 7820 [Note] InnoDB: Completed initialization of buffer pool
2017-11-28 19:42:44 7820 [Note] InnoDB: Highest supported file format is Barracuda.
2017-11-28 19:42:44 7820 [Note] InnoDB: Starting crash recovery from checkpoint LSN=556718604758
2017-11-28 19:42:52 7820 [Note] InnoDB: 128 out of 128 rollback segments are active.
2017-11-28 19:42:52 7820 [Note] InnoDB: Removed temporary tablespace data file: "ibtmp1"
2017-11-28 19:42:52 7820 [Note] InnoDB: Creating shared tablespace for temporary tables
2017-11-28 19:42:52 7820 [Note] InnoDB: Setting file '.\ibtmp1' size to 12 MB. Physically writing the file full; Please wait ...
2017-11-28 19:42:52 7820 [Note] InnoDB: File '.\ibtmp1' size is now 12 MB.
2017-11-28 19:42:52 7820 [Note] InnoDB: 5.7.14 started; log sequence number 556718604767
2017-11-28 19:42:52 4868 [Note] InnoDB: page_cleaner: 1000ms intended loop took 7784ms. The settings might not be optimal. (flushed=0 and evicted=0, during the time.)
2017-11-28 19:42:52 7380 [Note] InnoDB: Loading buffer pool(s) from C:\Program Files\MariaDB 10.2\data\ib_buffer_pool
2017-11-28 19:42:52 7820 [Note] Server socket created on IP: '::'.
2017-11-28 19:42:52 7820 [Note] Reading of all Master_info entries succeded
2017-11-28 19:42:52 7820 [Note] Added new Master_info '' to hash table
2017-11-28 19:42:52 7820 [Note] C:\Program Files\MariaDB 10.2\bin\mysqld.exe: ready for connections.
Version: '10.2.6-MariaDB' socket: '' port: 3306 mariadb.org binary distribution
2017-11-28 19:44:08 7380 [Note] InnoDB: Buffer pool(s) load completed at 171128 19:44:08
2017-11-28 19:44:20 9820 [Warning] InnoDB: Retry attempts for reading partial data failed.
2017-11-28 19:44:20 9820 [ERROR] InnoDB: Tried to read 16384 bytes at offset 4947968, but was only able to read 0
2017-11-28 19:44:20 9820 [ERROR] InnoDB: File (unknown): 'read' returned OS error 0. Cannot continue operation
171128 19:44:20 [ERROR] mysqld got exception 0x80000003 ;
This could be because you hit a bug. It is also possible that this binary
or one of the libraries it was linked against is corrupt, improperly built,
or misconfigured. This error can also be caused by malfunctioning hardware.
To report this bug, see https://mariadb.com/kb/en/reporting-bugs
We will try our best to scrape up some info that will hopefully help
diagnose the problem, but since we have already crashed,
something is definitely wrong and this may fail.
Server version: 10.2.6-MariaDB
key_buffer_size=134217728
read_buffer_size=131072
max_used_connections=1
max_threads=65537
thread_count=7
It is possible that mysqld could use up to
key_buffer_size + (read_buffer_size + sort_buffer_size)*max_threads = 136026 K bytes of memory
Hope that's ok; if not, decrease some variables in the equation.
Thread pointer: 0x7d1b9fd8
Attempting backtrace. You can use the following information to find out
where mysqld died. If you see no messages after this, something went
terribly wrong...
mysqld.exe!my_parameter_handler()[my_init.c:259]
mysqld.exe!raise()[signal.cpp:516]
mysqld.exe!abort()[abort.cpp:71]
mysqld.exe!os_file_handle_error_cond_exit()[os0file.cc:5209]
mysqld.exe!os_file_read_page()[os0file.cc:5091]
mysqld.exe!os_file_read_func()[os0file.cc:5433]
mysqld.exe!fil_io()[fil0fil.cc:5436]
mysqld.exe!buf_read_page_low()[buf0rea.cc:179]
mysqld.exe!buf_read_page()[buf0rea.cc:436]
mysqld.exe!buf_page_get_gen()[buf0buf.cc:4267]
mysqld.exe!btr_cur_search_to_nth_level()[btr0cur.cc:1115]
mysqld.exe!btr_pcur_open_low()[btr0pcur.ic:457]
mysqld.exe!btr_pcur_open_on_user_rec_func()[btr0pcur.cc:597]
mysqld.exe!dict_load_foreign()[dict0load.cc:3334]
mysqld.exe!dict_load_foreigns()[dict0load.cc:3587]
mysqld.exe!dict_load_table_one()[dict0load.cc:2958]
mysqld.exe!dict_load_table()[dict0load.cc:2670]
mysqld.exe!dict_table_open_on_name()[dict0dict.cc:1174]
mysqld.exe!ha_innobase::open_dict_table()[ha_innodb.cc:6976]
mysqld.exe!ha_innobase::open()[ha_innodb.cc:6618]
mysqld.exe!handler::ha_open()[handler.cc:2507]
mysqld.exe!open_table_from_share()[table.cc:3278]
mysqld.exe!open_table()[sql_base.cc:1874]
mysqld.exe!open_and_process_table()[sql_base.cc:3409]
mysqld.exe!open_tables()[sql_base.cc:3926]
mysqld.exe!open_and_lock_tables()[sql_base.cc:4682]
mysqld.exe!execute_sqlcom_select()[sql_parse.cc:6352]
mysqld.exe!mysql_execute_command()[sql_parse.cc:3448]
mysqld.exe!mysql_parse()[sql_parse.cc:7879]
mysqld.exe!dispatch_command()[sql_parse.cc:1814]
mysqld.exe!do_command()[sql_parse.cc:1361]
mysqld.exe!threadpool_process_request()[threadpool_common.cc:346]
mysqld.exe!tp_callback()[threadpool_common.cc:192]
ntdll.dll!TpPostWork()
ntdll.dll!RtlRealSuccessor()
kernel32.dll!BaseThreadInitThunk()
ntdll.dll!RtlUserThreadStart()
Trying to get some variables.
Some pointers may be invalid and cause the dump to abort.
Query (0x78b38ff0): SELECT COUNT(*) FROM curr.patient
Connection ID (thread ID): 9
Status: NOT_KILLED
Optimizer switch: index_merge=on,index_merge_union=on,index_merge_sort_union=on,index_merge_intersection=on,index_merge_sort_intersection=off,engine_condition_pushdown=off,index_condition_pushdown=on,derived_merge=on,derived_with_keys=on,firstmatch=on,loosescan=on,materialization=on,in_to_exists=on,semijoin=on,partial_match_rowid_merge=on,partial_match_table_scan=on,subquery_cache=on,mrr=off,mrr_cost_based=off,mrr_sort_keys=off,outer_join_with_cache=on,semijoin_with_cache=on,join_cache_incremental=on,join_cache_hashed=on,join_cache_bka=on,optimize_join_buffer_size=off,table_elimination=on,extended_keys=on,exists_to_in=on,orderby_uses_equalities=on,condition_pushdown_for_derived=on
The manual page at http://dev.mysql.com/doc/mysql/en/crashing.html contains
information that should help you find out what is causing the crash.

Answer 1

Based on the stack trace, it seems to be the InnoDB system tablespace that is shorter than expected. When the function dict_load_foreigns() is accessing the InnoDB system table SYS_FOREIGN or SYS_FOREIGN_COLS, it is requesting a page that is not in the buffer pool. The page read request causes InnoDB to commit suicide, because the file is too short.
InnoDB notoriously does not report the problematic file name. We should refactor the I/O code in MariaDB at some point. In this case, we do know that the problem is in the InnoDB system tablespace, because the InnoDB internal SYS_ tables are located there.
There already exist some related bugs in the MariaDB tracker. I think that this scenario is already covered by these:
MDEV-13542 Crashing on a corrupted page is unhelpful (Yes, I copied the title of the ancient MySQL Bug#10132)
MDEV-11633 Make the InnoDB system tablespace optional (This is not going to happen soon, and the design is not finalized)
It would be interesting to know how the corruption occurred in the first place. Before MDEV-11556, InnoDB data file extension in MariaDB was not fully crash-safe. (MySQL does not contain this fix at all.)
Could it be that the files were copied at some point? A bug in the copy procedure? Or could the system tablespace have originally consisted of multiple files, but the server was started up with the wrong innodb_data_file_path so that the last file(s) were ignored? Everything would appear fine until a page in the ‘missing’ files is being accessed.
You might ask: How to work around this error? Unfortunately, I don’t think that there currently is any way to skip the read of the foreign key metadata. So, if the metadata tables are corrupted, in the worst case you will be unable to access any InnoDB tables. For this, I would welcome a MariaDB bug report.

Answer 2

I wanted to reinstall MariaDB. When I started deinstall/change of MaridaDB through the application wizard of Window 7 it asked if I wanted to Change/Repair/Remove. I decided to repair. After that MariaDB worked as usual. That is I could submit queries without losing connection.
Upgrade from MariaDB from 10.2.6 to 10.2.11 didn't help. Repair allowed me to be successful.
Lesson learned: Before asking SO do repair MariaDB.