The SQL layer never acquires metadata locks (MDL) on the tables
that the tables that DML statement accesses is modifying.
However, the storage engine must access the parent table in order to
ensure that the child table will not refer to a non-existing record
in the parent table.
During certain DDL operations, the InnoDB table metadata (dict_table_t)
may be be freed and rebuilt. This would cause a race condition with
a concurrent INSERT that is attempting to report a FOREIGN KEY violation.
We work around the insufficient MDL during DML by acquiring exclusive
InnoDB table locks on all child tables during DDL. To avoid deadlocks,
we will follow the following order of acquisition:
1. tables whose REFERENCES clauses point to the current table
2. the current table that is being subjected to DDL
3. mysql.innodb_table_stats
4. mysql.innodb_index_stats
5. the InnoDB dictionary tables (SYS_TABLES and so on)
6. exclusive dict_sys.latch
Many InnoDB data dictionary cache operations require that the
table name be copied so that it will be NUL terminated.
(For example, SYS_TABLES.NAME is not guaranteed to be NUL-terminated.)
dict_table_t::is_garbage_name(): Check if a name belongs to
the background drop table queue.
dict_check_if_system_table_exists(): Remove.
dict_sys_t::load_sys_tables(): Load the non-hard-coded system tables
SYS_FOREIGN, SYS_FOREIGN_COLS, SYS_VIRTUAL on startup.
dict_sys_t::create_or_check_sys_tables(): Replaces
dict_create_or_check_foreign_constraint_tables() and
dict_create_or_check_sys_virtual().
dict_sys_t::load_table(): Replaces dict_table_get_low()
and dict_load_table().
dict_sys_t::find_table(): Renamed from get_table().
dict_sys_t::sys_tables_exist(): Check whether all the non-hard-coded
tables SYS_FOREIGN, SYS_FOREIGN_COLS, SYS_VIRTUAL exist.
trx_t::has_stats_table_lock(): Moved to dict0stats.cc.
Some error messages will now report table names in the internal
databasename/tablename format, instead of `databasename`.`tablename`.
fts_drop_orphaned_tables() takes long time to remove the orphaned
FTS tables. In order to reduce the time, do the following:
- Traverse fil_system.space_list and construct a set of
table_id,index_id of all FTS_*.ibd tablespaces.
- Traverse the sys_indexes table and ignore the entry
from the above collection if it exist.
- Existing elements in the collection can be considered as
orphaned fts tables. construct the table name from
(table_id,index_id) and invoke fts_drop_tables().
- Removed DICT_TF2_FTS_AUX_HEX_NAME flag usage from upgrade.
- is_aux_table() in dict_table_t to check whether the given name
is fts auxiliary table
fts_space_set_t is a structure to store set of parent table id
and index id
- Remove unused FTS function in fts0fts.cc
- Remove the fulltext index in row_format_redundant test case.
Because it deals with the condition that SYS_TABLES does have
corrupted entry and valid entry exist in SYS_INDEXES.
This is joint work with Thirunarayanan Balathandayuthapani.
The MDL interface between InnoDB and the rest of the server
(in storage/innobase/dict/dict0dict.cc and in include/)
is my work, while most everything else is Thiru's.
The collection of InnoDB persistent statistics and the
defragmentation were not refactored to use MDL. They will
keep relying on lower-level interlocking with
fil_check_pending_operations().
The purge of transaction history and the background operations on
fulltext indexes will use MDL. We will revert
commit 2c4844c9e7
(MDEV-17813) because thanks to MDL, purge cannot conflict
with DDL operations anymore. For a similar reason, we will remove
the MDEV-16222 test case from gcol.innodb_virtual_debug_purge.
Purge is essentially replacing all use of the global dict_sys.latch
with MDL. Purge will skip the undo log records for tables whose names
start with #sql-ib or #sql2. Theoretically, such tables might
be renamed back to visible table names if TRUNCATE fails to
create a new table, or the final rename in ALTER TABLE...ALGORITHM=COPY
fails. In that case, purge could permanently leave some garbage
in the table. Such garbage will be tolerated; the table would not
be considered corrupted.
To avoid repeated MDL releases and acquisitions,
trx_purge_attach_undo_recs() will sort undo log records by table_id,
and purge_node_t will keep the MDL and table handle open for multiple
successive undo log records.
get_purge_table(): A new accessor, used during the purge of
history for indexed virtual columns. This interface should ideally
not exist at all.
thd_mdl_context(): Accessor of THD::mdl_context.
Wrapped in a new thd_mdl_service.
dict_get_db_name_len(): Define inline.
dict_acquire_mdl_shared(): Acquire explicit shared MDL on a table name
if needed.
dict_table_open_on_id(): Return MDL_ticket, if requested.
dict_table_close(): Release MDL ticket, if requested.
dict_fts_index_syncing(), dict_index_t::index_fts_syncing: Remove.
row_drop_table_for_mysql() no longer needs to check these, because
MDL guarantees that a fulltext index sync will not be in progress
while MDL_EXCLUSIVE is protecting a DDL operation.
dict_table_t::parse_name(): Parse the table name for acquiring MDL.
purge_node_t::undo_recs: Change the type to std::list<trx_purge_rec_t*>
(different container, and storing also roll_ptr).
purge_node_t: Add mdl_ticket, last_table_id, purge_thd, mdl_hold_recs
for acquiring MDL and for keeping the table open across multiple
undo log records.
purge_vcol_info_t, row_purge_store_vsec_cur(), row_purge_restore_vsec_cur():
Remove. We will acquire the MDL earlier.
purge_sys_t::heap: Added, for reading undo log records.
fts_sync_during_ddl(): Invoked during ALGORITHM=INPLACE operations
to ensure that fts_sync_table() will not conflict with MDL_EXCLUSIVE.
Uses fts_t::sync_message for bookkeeping.
Historically, InnoDB split the redo log into at least 2 files.
MDEV-12061 allowed the minimum to be innodb_log_files_in_group=1,
but it kept the default at innodb_log_files_in_group=2.
Because performance seems to be slightly better with only one log file,
and because implementing an append-only variant of the log would require
a single file, let us define the default to be 1, and have
innodb_log_file_size=96M, to retain the same default total size.
This allows one to run the test suite even if any of the following
options are changed:
- character-set-server
- collation-server
- join-cache-level
- log-basename
- max-allowed-packet
- optimizer-switch
- query-cache-size and query-cache-type
- skip-name-resolve
- table-definition-cache
- table-open-cache
- Some innodb options
etc
Changes:
- Don't print out the value of system variables as one can't depend on
them to being constants.
- Don't set global variables to 'default' as the default may not
be the same as the test was started with if there was an additional
option file. Instead save original value and reset it at end of test.
- Test that depends on the latin1 character set should include
default_charset.inc or set the character set to latin1
- Test that depends on the original optimizer switch, should include
default_optimizer_switch.inc
- Test that depends on the value of a specific system variable should
set it in the test (like optimizer_use_condition_selectivity)
- Split subselect3.test into subselect3.test and subselect3.inc to
make it easier to set and reset system variables.
- Added .opt files for test that required specfic options that could
be changed by external configuration files.
- Fixed result files in rockdsb & tokudb that had not been updated for
a while.
Added --skip-test-db option to mysql_install_db. If specified, no test
database created and relevant grants issued.
Removed --skip-auth-anonymous-user option of mysql_install_db. Now it is
covered by --skip-test-db.
Dropped some Debian patches that did the same.
Removed unused make_win_bin_dist.1, make_win_bin_dist and
mysql_install_db.pl.in.
Remove the SHARED_SPACE flag that was erroneously introduced in
MariaDB 10.2.2, and shift the SYS_TABLES.TYPE flags back to where
they were before MariaDB 10.2.2. While doing this, ensure that
tables created with affected MariaDB versions can be loaded,
and also ensure that tables created with MySQL 5.7 using the
TABLESPACE attribute cannot be loaded.
MariaDB 10.2.2 picked the SHARED_SPACE flag from MySQL 5.7,
shifting the MariaDB 10.1 flags PAGE_COMPRESSION, PAGE_COMPRESSION_LEVEL,
ATOMIC_WRITES by one bit. The SHARED_SPACE flag would always
be written as 0 by MariaDB, because MariaDB does not support
CREATE TABLESPACE or CREATE TABLE...TABLESPACE for InnoDB.
So, instead of the bits AALLLLCxxxxxxx we would have
AALLLLC0xxxxxxx if the table was created with MariaDB 10.2.2
to 10.2.6. (AA=ATOMIC_WRITES, LLLL=PAGE_COMPRESSION_LEVEL,
C=PAGE_COMPRESSED, xxxxxxx=7 bits that were not moved.)
PAGE_COMPRESSED=NO implies LLLLC=00000. That is not a problem.
If someone created a table in MariaDB 10.2.2 or 10.2.3 with
the attribute ATOMIC_WRITES=OFF (value 2; AA=10) and without
PAGE_COMPRESSED=YES or PAGE_COMPRESSION_LEVEL, the table should be
rejected. We ignore this problem, because it should be unlikely
for anyone to specify ATOMIC_WRITES=OFF, and because 10.2.2 and
10.2.2 were not mature releases. The value ATOMIC_WRITES=ON (1)
would be interpreted as ATOMIC_WRITES=OFF, but starting with
MariaDB 10.2.4 the ATOMIC_WRITES attribute is ignored.
PAGE_COMPRESSED=YES implies that PAGE_COMPRESSION_LEVEL be between
1 and 9 and that ROW_FORMAT be COMPACT or DYNAMIC. Thus, the affected
wrong bit pattern in SYS_TABLES.TYPE is of the form AALLLL10DB00001
where D signals the presence of a DATA DIRECTORY attribute and B is 1
for ROW_FORMAT=DYNAMIC and 0 for ROW_FORMAT=COMPACT. We must interpret
this bit pattern as AALLLL1DB00001 (discarding the extraneous 0 bit).
dict_sys_tables_rec_read(): Adjust the affected bit pattern when
reading the SYS_TABLES.TYPE column. In case of invalid flags,
report both SYS_TABLES.TYPE (after possible adjustment) and
SYS_TABLES.MIX_LEN.
dict_load_table_one(): Replace an unreachable condition on
!dict_tf2_is_valid() with a debug assertion. The flags will already
have been validated by dict_sys_tables_rec_read(); if that validation
fails, dict_load_table_low() will have failed.
fil_ibd_create(): Shorten an error message about a file pre-existing.
Datafile::validate_to_dd(): Clarify an error message about tablespace
flags mismatch.
ha_innobase::open(): Remove an unnecessary warning message.
dict_tf_is_valid(): Simplify and stricten the logic. Validate the
values of PAGE_COMPRESSION. Remove error log output; let the callers
handle that.
DICT_TF_BITS: Remove ATOMIC_WRITES, PAGE_ENCRYPTION, PAGE_ENCRYPTION_KEY.
The ATOMIC_WRITES is ignored once the SYS_TABLES.TYPE has been validated;
there is no need to store it in dict_table_t::flags. The PAGE_ENCRYPTION
and PAGE_ENCRYPTION_KEY are unused since MariaDB 10.1.4 (the GA release
was 10.1.8).
DICT_TF_BIT_MASK: Remove (unused).
FSP_FLAGS_MEM_ATOMIC_WRITES: Remove (the flags are never read).
row_import_read_v1(): Display an error if dict_tf_is_valid() fails.
innodb.row_format_redundant: Really corrupt the SYS_TABLES.MIX_LEN,
and do not use any debug instrumentation. For tables created in the
system tablespace, the contents of the column will be ignored.
Only the table t1 will refuse to load.
dict_load_table_one(): Remove the DBUG_EXECUTE_IF instrumentation.
Omit a redundant error message "incorrect flags in SYS_TABLES".
dict_sys_tables_rec_read(): Partially revert the Oracle Bug#21644827
fix, and always report errors by the return value.
fts_create_in_mem_aux_table(): Do not rely on dict_table_t::flags2,
but instead evaluate the tablespace ID.
DICT_TF2_BITS: Reduce to the correct value of 7. The two extra
high-order bits were specific to MySQL 5.7.
srv_start_state_t: Document the flags. Replace SRV_START_STATE_STAT
with SRV_START_STATE_REDO. The srv_bg_undo_sources replaces the
original use of SRV_START_STATE_STAT.
dict_stats_thread_started, buf_dump_thread_started,
buf_flush_page_cleaner_thread_started: Remove (unused).
srv_shutdown_all_bg_threads(): Always wait for the I/O threads
to exit, also in read-only mode.
os_thread_free(): Remove.
When a slow shutdown is performed soon after spawning some work for
background threads that can create or commit transactions, it is possible
that new transactions are started or committed after the purge has finished.
This is violating the specification of innodb_fast_shutdown=0, namely that
the purge must be completed. (None of the history of the recent transactions
would be purged.)
Also, it is possible that the purge threads would exit in slow shutdown
while there exist active transactions, such as recovered incomplete
transactions that are being rolled back. Thus, the slow shutdown could
fail to purge some undo log that becomes purgeable after the transaction
commit or rollback.
srv_undo_sources: A flag that indicates if undo log can be generated
or the persistent, whether by background threads or by user SQL.
Even when this flag is clear, active transactions that already exist
in the system may be committed or rolled back.
innodb_shutdown(): Renamed from innobase_shutdown_for_mysql().
Do not return an error code; the operation never fails.
Clear the srv_undo_sources flag, and also ensure that the background
DROP TABLE queue is empty.
srv_purge_should_exit(): Do not allow the purge to exit if
srv_undo_sources are active or the background DROP TABLE queue is not
empty, or in slow shutdown, if any active transactions exist
(and are being rolled back).
srv_purge_coordinator_thread(): Remove some previous workarounds
for this bug.
innobase_start_or_create_for_mysql(): Set buf_page_cleaner_is_active
and srv_dict_stats_thread_active directly. Set srv_undo_sources before
starting the purge subsystem, to prevent immediate shutdown of the purge.
Create dict_stats_thread and fts_optimize_thread immediately
after setting srv_undo_sources, so that shutdown can use this flag to
determine if these subsystems were started.
dict_stats_shutdown(): Shut down dict_stats_thread. Backported from 10.2.
srv_shutdown_table_bg_threads(): Remove (unused).
