don't set vers_write=false if one vers column was used explicitly,
instead do vers_update_fields() for columns that do not have explicit
value. So, if row_start has the value and row_end not, row_end will
get max by default.
* clarify the help text for --system-versioning-insert-history
* move the vers_write=false check from Item_field::fix_fields()
next to other vers field checks in find_field_in_table()
* move row_start validation from handler::write_row() next to
vers_update_fields()
* make secure_timestamp check to happen in one place only,
extract it into a function is_set_timestamp_vorbidden().
* overwriting vers fields is an error, just like setting @@timestamp
* don't run vers_insert_history() for every row
1. system_versioning_insert_history session variable allows
pseudocolumns ROW_START and ROW_END be specified in INSERT,
INSERT..SELECT and LOAD DATA.
2. Cleaned up select_insert::send_data() from setting vers_write as
this parameter is now set on TABLE initialization.
4. Replication of system_versioning_insert_history via option_bits in
OPTIONS_WRITTEN_TO_BIN_LOG.
Underlying causes of all bugs mentioned below are same. This patch fixes
all of them:
1) MDEV-25028: ASAN use-after-poison in
base_list_iterator::next or Assertion `sl->join == 0' upon
INSERT .. RETURNING via PS
2) MDEV-25187: Assertion `inited == NONE || table->open_by_handler'
failed or Direct leak in init_dynamic_array2 upon INSERT .. RETURNING
and memory leak in init_dynamic_array2
3) MDEV-28740: crash in INSERT RETURNING subquery in prepared statements
4) MDEV-27165: crash in base_list_iterator::next
5) MDEV-29686: Assertion `slave == 0' failed in
st_select_lex_node::attach_single
Analysis:
consider this statement:
INSERT(1)...SELECT(2)...(SELECT(3)...) RETURNING (SELECT(4)...)
When RETURNING is encountered, add_slave() changes how selects are linked.
It makes the builtin_select(1) slave of SELECT(2). This causes
losing of already existing slave(3) (which is nested select of SELECT of
INSERT...SELECT). When really, builtin_select (1) shouldn't be slave to
SELECT(2) because it is not nested within it. Also, push_select() to use
correct context also changed how select are linked.
During reinit_stmt_before_use(), we expect the selects to
be cleaned-up and have join=0. Since these selects are not linked correctly,
clean-up doesn't happen correctly so join is not NULL. Hence the crash.
Fix:
IF we are parsing RETURNING, make is_parsing_returning= true for
current select. get rid of add_slave(). In place of push_select(), used
push_context() to have correct context (the context of builtin_select)
to resolve items in item_list. And add these items to item_list of
builtin_select.
The problem is that if table definition cache (TDC) is full of real tables
which are in tables cache, view definition can not stay there so will be
removed by its own underlying tables.
In situation above old mechanism of detection matching definition in PS
and current version always require reprepare and so prevent executing
the PS.
One work around is to increase TDC, other - improve version check for
views/triggers (which is done here). Now in suspicious cases we check:
- timestamp (microseconds) of the view to be sure that version really
have changed;
- time (microseconds) of creation of a trigger related to time
(microseconds) of statement preparation.
- Added missing information about database of corresponding table for various types of commands
- Update some typos
- Reviewed by: <vicentiu@mariadb.org>
Atomic CREATE OR REPLACE allows to keep an old table intact if the
command fails or during the crash. That is done through creating
a table with a temporary name and filling it with the data
(for CREATE OR REPLACE .. SELECT), then renaming the original table
to another temporary (backup) name and renaming the replacement table
to original table. The backup table is kept until the last chance of
failure and if that happens, the replacement table is thrown off and
backup recovered. When the command is complete and logged the backup
table is deleted.
Atomic replace algorithm
Two DDL chains are used for CREATE OR REPLACE:
ddl_log_state_create (C) and ddl_log_state_rm (D).
1. (C) Log CREATE_TABLE_ACTION of TMP table (drops TMP table);
2. Create new table as TMP;
3. Do everything with TMP (like insert data);
finalize_atomic_replace():
4. Link chains: (D) is executed only if (C) is closed;
5. (D) Log DROP_ACTION of BACKUP;
6. (C) Log RENAME_TABLE_ACTION from ORIG to BACKUP (replays BACKUP -> ORIG);
7. Rename ORIG to BACKUP;
8. (C) Log CREATE_TABLE_ACTION of ORIG (drops ORIG);
9. Rename TMP to ORIG;
finalize_ddl() in case of success:
10. Close (C);
11. Replay (D): BACKUP is dropped.
finalize_ddl() in case of error:
10. Close (D);
11. Replay (C):
1) ORIG is dropped (only after finalize_atomic_replace());
2) BACKUP renamed to ORIG (only after finalize_atomic_replace());
3) drop TMP.
If crash happens (C) or (D) is replayed in reverse order. (C) is
replayed if crash happens before it is closed, otherwise (D) is
replayed.
Temporary table for CREATE OR REPLACE
Before dropping "old" table, CREATE OR REPLACE creates "tmp" table.
ddl_log_state_create holds the drop of the "tmp" table. When
everything is OK (data is inserted, "tmp" is ready) ddl_log_state_rm
is written to replace "old" with "tmp". Until ddl_log_state_create
is closed ddl_log_state_rm is not executed.
After the binlogging is done ddl_log_state_create is closed. At that
point ddl_log_state_rm is executed and "tmp" is replaced with
"old". That is: final rename is done by the DDL log.
With that important role of DDL log for CREATE OR REPLACE operation
replay of ddl_log_state_rm must fail at the first hit error and
print the error message if possible. F.ex. foreign key error is
discovered at this phase: InnoDB rejects to drop the "old" table and
returns corresponding foreign key error code.
Additional notes
- CREATE TABLE without REPLACE is not affected by this commit.
- Engines having HTON_EXPENSIVE_RENAME flag set are not affected by
this commit.
- CREATE TABLE .. SELECT XID usage is fixed and now there is no need
to log DROP TABLE via DDL_CREATE_TABLE_PHASE_LOG (see comments in
do_postlock()). XID is now correctly updated so it disables
DDL_LOG_DROP_TABLE_ACTION. Note that binary log is flushed at the
final stage when the table is ready. So if we have XID in the
binary log we don't need to drop the table.
- Three variations of CREATE OR REPLACE handled:
1. CREATE OR REPLACE TABLE t1 (..);
2. CREATE OR REPLACE TABLE t1 LIKE t2;
3. CREATE OR REPLACE TABLE t1 SELECT ..;
- Test case uses 6 combinations for engines (aria, aria_notrans,
myisam, ib, lock_tables, expensive_rename) and 2 combinations for
binlog types (row, stmt). Combinations help to check differences
between the results. Error failures are tested for the above three
variations.
- expensive_rename tests CREATE OR REPLACE without atomic
replace. The effect should be the same as with the old behaviour
before this commit.
- Triggers mechanism is unaffected by this change. This is tested in
create_replace.test.
- LOCK TABLES is affected. Lock restoration must be done after "rm"
chain is replayed.
- Moved ddl_log_complete() from send_eof() to finalize_ddl(). This
checkpoint was not executed before for normal CREATE TABLE but is
executed now.
- CREATE TABLE will now rollback also if writing to the binary
logging failed. See rpl_gtid_strict.test
Rename and drop via DDL log
We replay ddl_log_state_rm to drop the old table and rename the
temporary table. In that case we must throw the correct error
message if ddl_log_revert() fails (f.ex. on FK error).
If table is deleted earlier and not via DDL log and the crash
happened, the create chain is not closed. Linked drop chain is not
executed and the new table is not installed. But the old table is
already deleted.
ddl_log.cc changes
Now we can place action before DDL_LOG_DROP_INIT_ACTION and it will
be replayed after DDL_LOG_DROP_TABLE_ACTION.
report_error parameter for ddl_log_revert() allows to fail at first
error and print the error message if possible.
ddl_log_execute_action() now can print error message.
Since we now can handle errors from ddl_log_execute_action() (in
case of non-recovery execution) unconditional setting "error= TRUE"
is wrong (it was wrong anyway because it was overwritten at the end
of the function).
On XID usage
Like with all other atomic DDL operations XID is used to avoid
inconsistency between master and slave in the case of a crash after
binary log is written and before ddl_log_state_create is closed. On
recovery XIDs are taken from binary log and corresponding DDL log
events get disabled. That is done by
ddl_log_close_binlogged_events().
On linking two chains together
Chains are executed in the ascending order of entry_pos of execute
entries. But entry_pos assignment order is undefined: it may assign
bigger number for the first chain and then smaller number for the
second chain. So the execution order in that case will be reverse:
second chain will be executed first.
To avoid that we link one chain to another. While the base chain
(ddl_log_state_create) is active the secondary chain
(ddl_log_state_rm) is not executed. That is: only one chain can be
executed in two linked chains.
The interface ddl_log_link_chains() was done in "MDEV-22166
ddl_log_write_execute_entry() extension".
More on CREATE OR REPLACE .. SELECT
We use create_and_open_tmp_table() like in ALTER TABLE to create
temporary TABLE object (tmp_table is (NON_)TRANSACTIONAL_TMP_TABLE).
After we created such TABLE object we use create_info->tmp_table()
instead of table->s->tmp_table when we need to check for
parser-requested tmp-table.
External locking is required for temporary table created by
create_and_open_tmp_table(). F.ex. that disables logging for Aria
transactional tables and without that (when no mysql_lock_tables()
is done) it cannot work correctly.
For making external lock the patch requires Aria table to work in
non-transactional mode. That is usually done by
ha_enable_transaction(false). But we cannot disable transaction
completely because: 1. binlog rollback removes pending row events
(binlog_remove_pending_rows_event()). The row events are added
during CREATE .. SELECT data insertion phase. 2. replication slave
highly depends on transaction and cannot work without it.
So we put temporary Aria table into non-transactional mode with
"thd->transaction->on hack". See comment for on_save variable.
Note that Aria table has internal_table mode. But we cannot use it
because:
if (!internal_table)
{
mysql_mutex_lock(&THR_LOCK_myisam);
old_info= test_if_reopen(name_buff);
}
For internal_table test_if_reopen() is not called and we get a new
MARIA_SHARE for each file handler. In that case duplicate errors are
missed because insert and lookup in CREATE .. SELECT is done via two
different handlers (see create_lookup_handler()).
For temporary table before dropping TABLE_SHARE by
drop_temporary_table() we must do ha_reset(). ha_reset() releases
storage share. Without that the share is kept and the second CREATE
OR REPLACE .. SELECT fails with:
HA_ERR_TABLE_EXIST (156): MyISAM table '#sql-create-b5377-4-t2' is
in use (most likely by a MERGE table). Try FLUSH TABLES.
HA_EXTRA_PREPARE_FOR_DROP also removes MYISAM_SHARE, but that is
not needed as ha_reset() does the job.
ha_reset() is usually done by
mark_tmp_table_as_free_for_reuse(). But we don't need that mechanism
for our temporary table.
Atomic_info in HA_CREATE_INFO
Many functions in CREATE TABLE pass the same parameters. These
parameters are part of table creation info and should be in
HA_CREATE_INFO (or whatever). Passing parameters via single
structure is much easier for adding new data and
refactoring.
InnoDB changes (revised by Marko Mäkelä)
row_rename_table_for_mysql(): Specify the treatment of FOREIGN KEY
constraints in a 4-valued enum parameter. In cases where FOREIGN KEY
constraints cannot exist (partitioned tables, or internal tables of
FULLTEXT INDEX), we can use the mode RENAME_IGNORE_FK.
The mod RENAME_REBUILD is for any DDL operation that rebuilds the
table inside InnoDB, such as TRUNCATE and native ALTER TABLE
(or OPTIMIZE TABLE). The mode RENAME_ALTER_COPY is used solely
during non-native ALTER TABLE in ha_innobase::rename_table().
Normal ha_innobase::rename_table() will use the mode RENAME_FK.
CREATE OR REPLACE will rename the old table (if one exists) along
with its FOREIGN KEY constraints into a temporary name. The replacement
table will be initially created with another temporary name.
Unlike in ALTER TABLE, all FOREIGN KEY constraints must be renamed
and not inherited as part of these operations, using the mode RENAME_FK.
dict_get_referenced_table(): Let the callers convert names when needed.
create_table_info_t::create_foreign_keys(): CREATE OR REPLACE creates
the replacement table with a temporary name table, so for
self-references foreign->referenced_table will be a table with
temporary name and charset conversion must be skipped for it.
Reviewed by:
Michael Widenius <monty@mariadb.org>
1. For INSERT..SELECT statements: don't include table/view the data
is inserted into in the list of leaf tables
2. Remove duplicated and dead code related to table_count
DBUG_ASSERT(0) was added by MDEV-17554 (auto-create history
partitions) as an experimental measure. Testing has shown this
conditional branch of can_recover_from_failed_open() can be possible
due to MDL deadlock.
The fix replaces DBUG_ASSERT with more specific one for
!OT_ADD_HISTORY_PARTITION.
Test case was synchronized to reproduce deadlock always and commented
with execution path of MDL locking for Good (no deadlock) and Bad
(deadlock). The logging was done with the help of preceding patch for
debug MDL tracing.
The issue was that flush_tables() didn't take a MDL lock on cached
TABLE_SHARE before calling open_table() to do a HA_EXTRA_FLUSH call.
Most engines seams to have no issue with it, but apparantly this conflicts
with InnoDB in 10.6 when using TRUNCATE
Fixed by taking a MDL lock before trying to open the table in
flush_tables().
There is no test case as it hard to repeat the scheduling that causes
the error. I did run the test case in MDEV-28897 to verify
that the bug is fixed.
If UPDATE/DELETE does not change data it is skipped from
replication. We now force replication of such events when they trigger
partition auto-creation.
For ROLLBACK it is as simple as set OPTION_KEEP_LOG
flag. trans_cannot_safely_rollback() does the rest.
For UPDATE/DELETE .. LIMIT 0 we make additional binlog_query() calls
at the early points of return.
As a safety measure we also convert row format into statement if it is
needed. The condition is decided by
binlog_need_stmt_format(). Basically if there are some row events in
cache we don't need that: table open of row event will trigger
auto-creation anyway.
Multi-update/delete works via mysql_select(). There is no early points
of return, so binlogging is always checked by
send_eof()/abort_resultset(). But we must comply with the above
measure of converting into statement.
:: Syntax change ::
Keyword AUTO enables history partition auto-creation.
Examples:
CREATE TABLE t1 (x int) WITH SYSTEM VERSIONING
PARTITION BY SYSTEM_TIME INTERVAL 1 HOUR AUTO;
CREATE TABLE t1 (x int) WITH SYSTEM VERSIONING
PARTITION BY SYSTEM_TIME INTERVAL 1 MONTH
STARTS '2021-01-01 00:00:00' AUTO PARTITIONS 12;
CREATE TABLE t1 (x int) WITH SYSTEM VERSIONING
PARTITION BY SYSTEM_TIME LIMIT 1000 AUTO;
Or with explicit partitions:
CREATE TABLE t1 (x int) WITH SYSTEM VERSIONING
PARTITION BY SYSTEM_TIME INTERVAL 1 HOUR AUTO
(PARTITION p0 HISTORY, PARTITION pn CURRENT);
To disable or enable auto-creation one can use ALTER TABLE by adding
or removing AUTO from partitioning specification:
CREATE TABLE t1 (x int) WITH SYSTEM VERSIONING
PARTITION BY SYSTEM_TIME INTERVAL 1 HOUR AUTO;
# Disables auto-creation:
ALTER TABLE t1 PARTITION BY SYSTEM_TIME INTERVAL 1 HOUR;
# Enables auto-creation:
ALTER TABLE t1 PARTITION BY SYSTEM_TIME INTERVAL 1 HOUR AUTO;
If the rest of partitioning specification is identical to CREATE TABLE
no repartitioning will be done (for details see MDEV-27328).
:: Description ::
Before executing history-generating DML command (see the list of commands below)
add N history partitions, so that N would be sufficient for potentially
generated history. N > 1 may be required when history partitions are switched
by INTERVAL and current_timestamp is N times further than the interval
boundary of the last history partition.
If the last history partition equals or exceeds LIMIT records then new history
partition is created and selected as the working partition. According to
MDEV-28411 partitions cannot be switched (or created) while the command is
running. Thus LIMIT does not carry strict limitation and the history partition
size must be planned as LIMIT value plus average number of history one DML
command can generate.
Auto-creation is implemented by synchronous fast_alter_partition_table() call
from the thread of the executed DML command before the command itself is run
(by the fallback and retry mechanism similar to Discovery feature,
see Open_table_context).
The name for newly added partitions are generated like default partition names
with extension of MDEV-22155 (which avoids name clashes by extending assignment
counter to next free-enough gap).
These DML commands can trigger auto-creation:
DELETE (including multitable DELETE, excluding DELETE HISTORY)
UPDATE (including multitable UPDATE)
REPLACE (including REPLACE .. SELECT)
INSERT .. ON DUPLICATE KEY UPDATE (including INSERT .. SELECT .. ODKU)
LOAD DATA .. REPLACE
:: Bug fixes ::
MDEV-23642 Locking timeout caused by auto-creation affects original DML
The reasons for this are:
- Do not disrupt main business process (the history is auxiliary service);
- Consequences are non-fatal (history is not lost, but comes into wrong
partition; fixed by partitioning rebuild);
- There is more freedom for application to fail in this case or not: it may
read warning info and find corresponding error number.
- While non-failing command is easy to handle by an application and fail it,
the opposite is hard to handle: there is no automatic actions to fix
failed command and retry, DBA intervention is required and until then
application is non-functioning.
MDEV-23639 Auto-create does not work under LOCK TABLES or inside triggers
Don't do tdc_remove_table() for OT_ADD_HISTORY_PARTITION because it is
not possible in locked tables mode.
LTM_LOCK_TABLES mode (and LTM_PRELOCKED_UNDER_LOCK_TABLES) works out
of the box as fast_alter_partition_table() can reopen tables via
locked_tables_list.
In LTM_PRELOCKED we reopen and relock table manually.
:: More fixes ::
* some_table_marked_for_reopen flag fix
some_table_marked_for_reopen affets only reopen of
m_locked_tables. I.e. Locked_tables_list::reopen_tables() reopens only
tables from m_locked_tables.
* Unused can_recover_from_failed_open() condition
Is recover_from_failed_open() can be really used after
open_and_process_routine()?
:: Reviewed by ::
Sergei Golubchik <serg@mariadb.org>
Moved LIMIT warning from vers_set_hist_part() to new call
vers_check_limit() at table unlock phase. At that point
read_partitions bitmap is already pruned by DML code (see
prune_partitions(), find_used_partitions()) so we have to set
corresponding bits for working history partition.
Also we don't do my_error(ME_WARNING|ME_ERROR_LOG), because at that
point it doesn't update warnings number, so command reports 0 warnings
(but warning list is still updated). Instead we do
push_warning_printf() and sql_print_warning() separately.
Under LOCK TABLES external_lock(F_UNLCK) is not executed. There is
start_stmt(), but no corresponding "stop_stmt()". So for that mode we
call vers_check_limit() directly from close_thread_tables().
Test result has been changed according to new LIMIT and warning
printing algorithm. For convenience all LIMIT warnings are marked with
"You see warning above ^".
TODO MDEV-20345 fixed. Now vers_history_generating() contains
fine-grained list of DML-commands that can generate history (and TODO
mechanism worked well).
- Describe the lifetime of EXPLAIN data structures in
sql_explain.h:ExplainDataStructureLifetime.
- Make Item_field::set_field() call set_refers_to_temp_table()
when it refers to a temp. table.
- Introduce QT_DONT_ACCESS_TMP_TABLES flag for Item::print.
It directs Item_field::print to not try access its the
temp table.
- Introduce Explain_query::notify_tables_are_closed()
and call it right before the query closes its tables.
- Make Explain data stuctures' print_explain_json() methods
accept "no_tmp_tbl" parameter which means pass
QT_DONT_ACCESS_TMP_TABLES when printing items.
- Make Show_explain_request::call_in_target_thread() not call
set_current_thd(). This wasn't needed as the code inside
lex->print_explain() uses output->thd anyway. output->thd
refers to the SHOW command's THD object.
EXPLAIN FOR CONNECTION is a MySQL-compatible syntax for SHOW EXPLAIN.
This commit also adds support for FORMAT=JSON to SHOW EXPLAIN,
so the possible options to get JSON-formatted output are:
- SHOW EXPLAIN FORMAT=JSON FOR $con
- EXPLAIN FORMAT=JSON FOR CONNECTION $con
column generated using date_format() and if()
vcol_info->expr is allocated on expr_arena at parsing stage. Since
expr item is allocated on expr_arena all its containee items must be
allocated on expr_arena too. Otherwise fix_session_expr() will
encounter prematurely freed item.
When table is reopened from cache vcol_info contains stale
expression. We refresh expression via TABLE::vcol_fix_exprs() but
first we must prepare a proper context (Vcol_expr_context) which meets
some requirements:
1. As noted above expr update must be done on expr_arena as there may
be new items created. It was a bug in fix_session_expr_for_read() and
was just not reproduced because of no second refix. Now refix is done
for more cases so it does reproduce. Tests affected: vcol.binlog
2. Also name resolution context must be narrowed to the single table.
Tested by: vcol.update main.default vcol.vcol_syntax gcol.gcol_bugfixes
3. sql_mode must be clean and not fail expr update.
sql_mode such as MODE_NO_BACKSLASH_ESCAPES, MODE_NO_ZERO_IN_DATE, etc
must not affect vcol expression update. If the table was created
successfully any further evaluation must not fail. Tests affected:
main.func_like
Reviewed by: Sergei Golubchik <serg@mariadb.org>
1. moved fix_vcol_exprs() call to open_table()
mysql_alter_table() doesn't do lock_tables() so it cannot win from
fix_vcol_exprs() from there. Tests affected: main.default_session
2. Vanilla cleanups and comments.
