Problem
========
On a parallel, delayed replica, Seconds_Behind_Master will not be
calculated until after MASTER_DELAY seconds have passed and the
event has finished executing, resulting in potentially very large
values of Seconds_Behind_Master (which could be much larger than the
MASTER_DELAY parameter) for the entire duration the event is
delayed. This contradicts the documented MASTER_DELAY behavior,
which specifies how many seconds to withhold replicated events from
execution.
Solution
========
After a parallel replica idles, the first event after idling should
immediately update last_master_timestamp with the time that it began
execution on the primary.
Reviewed By
===========
Andrei Elkin <andrei.elkin@mariadb.com>
This patch fixes the patch for bug MDEV-30248 that unsatisfactorily
resolved the problem of resolution of references to CTE. In some cases
when such a reference has the same table name as the name of one of
CTEs containing this reference the reference could be resolved incorrectly
that led to an invalid select tree where units could be mutually dependent.
This in its turn could lead to an infinite sequence of recursive calls or
to falls into infinite loops.
The patch also removes LEX::resolve_references_to_cte_in_hanging_cte() as
with the new code for resolution of CTE references the call of this
function is not needed anymore.
Approved by Oleksandr Byelkin <sanja@mariadb.com>
(Initial patch by Varun Gupta. Amended and added comments).
When the query has both
1. Aggregate functions that require sorting data by group, and
2. Window functions
we need to use two temporary tables. The first temp.table will hold the
join output. Then it is passed to filesort(). Reading it in sorted
order allows to compute the aggregate functions.
Then, we need to write their values into the second temp. table. Then,
Window Function computation step can pass that to filesort() and read
them in the order it needs.
Failure to create the second temp. table would cause an assertion
failure: window function could would not find where to get the values
of the aggregate functions.
disable bulk insert optimization if long uniques are used, because they
need to read the table (index_read) after every inserted now. And bulk
insert optimization might disable indexes.
bulk insert is already disabled in other cases when there are chances
that the table will be read duing the bulk insert.
plugin_vars_free_values() was walking plugin sysvars and thus
did not free memory of plugin PLUGIN_VAR_NOSYSVAR vars.
* change it to walk all plugin vars
* add the pluginname_ prefix to NOSYSVARS var names too,
so that plugin_vars_free_values() would be able to find their
bookmarks
Use SELECT_LEX to save lists for ORDER BY and GROUP BY before parsing
WINDOW clauses / specifications. This is needed for proper parsing
of a nested WINDOW clause when a WINDOW clause is used in a subquery
contained in another WINDOW clause.
Fix assignment of empty SQL_I_List to another one (in case of empty list
next shoud point on first).
Item_singlerow_subselect may be converted to Item_cond during
optimization. So there is a possibility of constructing nested
Item_cond_and or Item_cond_or which is not allowed (such
conditions must be flattened).
This commit checks if such kind of optimization has been applied
and flattens the condition if needed
This commit contains only a mtr test for reproducing the issue in MDEV-29512
The actual fix will be pushed in wsrep-lib repository
The hanging in MDEV-29512 happens when binlog purging is attempted, and there is
one local BF aborted transaction waiting for commit monitor.
The test will launch two node cluster and enable binlogging with expire log days,
to force binlog purging to happen.
A local transaction is executed so that will become BF abort victim, and has advanced
to replication stage waiting for commit monitor for final cleanup (to mark position in innodb)
after that, applier is released to complete the BF abort and due to binlog configuration,
starting the binlog purging. This is where the hanging would occur, if code is buggy
Reviewed-by: Jan Lindström <jan.lindstrom@mariadb.com>
Created mtr test for reproducing the crash
Developed actual fix for the issue.
Setting THD::system_thread_info.rpl_sql_info for replayer thread,
same way as it is handled for appliers.
Recorded test result, with the fix
Reviewed-by: Jan Lindström <jan.lindstrom@mariadb.com>
node->is_delete was incorrectly set to NO_DELETE for a set of operations.
In general we shouldn't rely on sql_command and look for more abstract ways
to control the behavior.
trg_event_map seems to be a suitable way. To mind replica nodes, it is ORed
with slave_fk_event_map, which stores trg_event_map when replica has
triggers disabled.
Problem:
=======
Mysqlbinlog cannot show the type of a compressed
column when two levels of verbosity is provided.
Solution:
========
Extend the log event printing logic to handle and
tag compressed types.
Behavioral Changes:
==================
Old: When mysqlbinlog is called in verbose mode and
the database uses compressed columns, an error is
returned to the user.
New: The output will append “ COMPRESSED” on the
type of compressed columns
Reviewed By
===========
Andrei Elkin <andrei.elkin@mariadb.com>
(Variant 3, initial variant was by Rex Jonston)
A LEFT JOIN with a constant as a column of the inner table produced wrong
query result if the optimizer had to write the inner table column into a
temp table. Query pattern:
SELECT ...
FROM (SELECT /*non-mergeable select*/
FROM t1 LEFT JOIN (SELECT 'Y' as Val) t2 ON ...) as tbl
Fixed this by adding Item_direct_view_ref::save_in_field() which follows
the pattern of Item_direct_view_ref's save_org_in_field(),
save_in_result_field() and val_XXX() functions:
* call check_null_ref() and handle NULL value
* if we didn't get a NULL-complemented row, call Item_direct_ref's function.
it's incorrect to use change_item_tree() to replace arguments
of top-level AND/OR, because they (arguments) are stored in a List,
so a pointer to an argument is in the list_node, and individual
list_node's of top-level AND/OR can be deleted in Item_cond::build_equal_items().
In that case rollback_item_tree_changes() will modify the deleted object.
Luckily, it's not needed to use change_item_tree() for top-level
AND/OR, because the whole top-level item is copied and preserved
in prep_where and prep_on, and restored from there.
So, just don't.
Additionally to the test case in the commit it fixes
* ASAN failure of main.opt_tvc --ps
* ASAN failure of main.having_cond_pushdown --ps
when an internal temporary table field is created from a real field,
a new temp field should only copy a default from the source field
when the latter has it
when creating a temp table field from an actual table field,
these two fields are supposed to be mostly identical
(except for BIT field storage), in particular, temp field should
have the same default as the orig field, even if the sql_mode has
been changed meanwhile (e.g. to include NO_ZERO_DATE)
regression from MDEV-29540 / 8c38939369.
INSERT SELECT errors needed to be unconditionally ignored.
As this touches the CREATE .. SELECT functionality, show
the equalivent test there.
This bug affected queries with nested left joins having the same last inner
table such that not_exists optimization could be applied to the most inner
outer join when optimizer chose to use join buffers. The bug could lead to
producing wrong a result set.
If the WHERE condition a query contains a conjunctive IS NULL predicate
over a non-nullable column of an inner table of a not nested outer join
then not_exists optimization can be applied to tho the outer join. With
this optimization when looking for matches for a certain record from the
outer table of the join the records of the inner table can be ignored
right after the first match satisfying the ON condition is found.
In the case of nested outer joins having the same last inner table this
optimization still can be applied but only if all ON conditions of the
embedding outer joins are satisfied. Such check was missing in the code
that tried to apply not_exists optimization when join buffers were used
for outer join operations.
This problem has been already fixed in the patch for bug MDEV-7992. Yet
there it was resolved only for the cases when join buffers were not used
for outer joins.
Approved by Oleksandr Byelkin <sanja@mariadb.com>
the parser couldn't parse `1=2 not between 3 and 5`
after `2` it expected only NOT2_SYM, but not NOT_SYM
(visible from the sql_yacc.output file), which resulted in
Syntax error ... near 'not between 3 and 4'
The parser was confused by a rather low NOT_SYM precedence and
%prec BETWEEN_SYM didn't resolve this confusion.
As a fix, let's remove any %precedence from NOT_SYM and
specify %prec explicitly in the only place where it matters for NOT_SYM.
In other places, such as for NOT BETWEEN, NOT_SYM won't have a
precedence, so bison won't be confused about it.
The idea is to put Item_direct_ref_to_item as a transparent and
permanent wrapper before a string which require conversion.
So that Item_direct_ref_to_item would be the only place where
the pointer to the string item is stored, this pointer can be changed
and restored during PS execution as needed. And if any permanent
(subquery) optimization would need a pointer to the item,
it'll use a pointer to the Item_direct_ref_to_item - which is
a permanent item and won't go away.
1. In case of system-versioned table add row_end into FTS_DOC_ID index
in fts_create_common_tables() and innobase_create_key_defs().
fts_n_uniq() returns 1 or 2 depending on whether the table is
system-versioned.
After this patch recreate of FTS_DOC_ID index is required for
existing system-versioned tables. If you see this message in error
log or server warnings: "InnoDB: Table db/t1 contains 2 indexes
inside InnoDB, which is different from the number of indexes 1
defined in the MariaDB" use this command to fix the table:
ALTER TABLE db.t1 FORCE;
2. Fix duplicate history for secondary unique index like it was done
in MDEV-23644 for clustered index (932ec586aa). In case of
existing history row which conflicts with currently inseted row we
check in row_ins_scan_sec_index_for_duplicate() whether that row
was inserted as part of current transaction. In that case we
indicate with DB_FOREIGN_DUPLICATE_KEY that new history row is not
needed and should be silently skipped.
3. Some parts of MDEV-21138 (7410ff436e) reverted. Skipping of
FTS_DOC_ID index for history rows made problems with purge
system. Now this is fixed differently by p.2.
4. wait_all_purged.inc checks that we didn't affect non-history rows
so they are deleted and purged correctly.
Additional FTS fixes
fts_init_get_doc_id(): exclude history rows from max_doc_id
calculation. fts_init_get_doc_id() callback is used only for crash
recovery.
fts_add_doc_by_id(): set max value for row_end field.
fts_read_stopword(): stopwords table can be system-versioned too. We
now read stopwords only for current data.
row_insert_for_mysql(): exclude history rows from doc_id validation.
row_merge_read_clustered_index(): exclude history_rows from doc_id
processing.
fts_load_user_stopword(): for versioned table retrieve row_end field
and skip history rows. For non-versioned table we retrieve 'value'
field twice (just for uniformity).
FTS tests for System Versioning now include maybe_versioning.inc which
adds 3 combinations:
'vers' for debug build sets sysvers_force and
sysvers_hide. sysvers_force makes every created table
system-versioned, sysvers_hide hides WITH SYSTEM VERSIONING
for SHOW CREATE.
Note: basic.test, stopword.test and versioning.test do not
require debug for 'vers' combination. This is controlled by
$modify_create_table in maybe_versioning.inc and these
tests run WITH SYSTEM VERSIONING explicitly which allows to
test 'vers' combination on non-debug builds.
'vers_trx' like 'vers' sets sysvers_force_trx and sysvers_hide. That
tests FTS with trx_id-based System Versioning.
'orig' works like before: no System Versioning is added, no debug is
required.
Upgrade/downgrade test for System Versioning is done by
innodb_fts.versioning. It has 2 combinations:
'prepare' makes binaries in std_data (requires old server and OLD_BINDIR).
It tests upgrade/downgrade against old server as well.
'upgrade' tests upgrade against binaries in std_data.
Cleanups:
Removed innodb-fts-stopword.test as it duplicates stopword.test
Adds new parameter $restart_bindir for restart_mysqld.inc.
Example:
let $restart_bindir= /home/midenok/src/mariadb/10.3b/build;
--source include/restart_mysqld.inc
It is good to return back original server before check_mysqld will be
run at the test end:
let $restart_bindir=;
--source include/restart_mysqld.inc
Before the fix next-key lock was requested only if a record was
delete-marked for locking unique search in RR isolation level.
There can be several delete-marked records for the same unique key,
that's why InnoDB scans the records until eighter non-delete-marked record
is reached or all delete-marked records with the same unique key are
scanned.
For range scan next-key locks are used for RR to protect scanned range from
inserting new records by other transactions. And this is the reason of why
next-key locks are used for delete-marked records for unique searches.
If a record is not delete-marked, the requested lock type was "not-gap".
When a record is not delete-marked during lock request by trx 1, and
some other transaction holds conflicting lock, trx 1 creates waiting
not-gap lock on the record and suspends. During trx 1 suspending the
record can be delete-marked. And when the lock is granted on conflicting
transaction commit or rollback, its type is still "not-gap". So we have
"not-gap" lock on delete-marked record for RR. And this let some other
transaction to insert some record with the same unique key when trx 1 is
not committed, what can cause isolation level violation.
The fix is to set next-key locks for both delete-marked and
non-delete-marked records for unique search in RR.
An unfortunate change to the default behavior of the handling of
core dumps was implemented in
commit e9be5428a2
by making MTR_PRINT_CORE=small the default value, that is, to
only display the stack trace of one thread in crash reports.
Many if not most failures that occur in regression tests are
sporadic and involve race conditions or deadlocks. To be able to
analyze such failures, having the stack traces of all active threads
is a must, because CI environments typically do not save any core dumps.
While the environment variable MTR_PRINT_CORE could be set in CI
environments to compensate for the unfortunate change, it is better
to revert to the old default (dumping all threads) so that
no explicit action will be required from maintainers of independent
CI systems. In that case, if something fails once in a blue moon,
we can have some hope of diagnosing it based on the output.
We fix this regression by defaulting the unset environment variable
MTR_PRINT_CORE to "medium".
Some tests drop the default mtr database "test". This may fail due
to the directory not being empty. InnoDB may not delete all tables
immediately, due to the "background drop table queue" or its
replacement in commit 1bd681c8b3
(the purge of history would clean up after a DDL operation during
which the server was killed).
Let us try to avoid "drop database test" whenever it is easily possible.
Where it is not, SET GLOBAL innodb_max_purge_lag_wait=0 will ensure
that the replacement of the "background drop table queue" will have
completed its job.
Consistent with MDEV-4206 and empty log_slow_filter still means
no explict filtering. Since 21518ab2e4 however the
log_queries_not_using_indexes became stored in the same variable.
As we need to test for the absense of log_queries_not_using_indexes
the SERVER_QUERY_NO_INDEX USED part of log_slow_statement, the empty
criteria resulted in an always true to log queries not using indexes if
log_slow_filter was set to empty.
Adjusted the log_slow.test for MDEV-4206 as slow_log_query has been
global and session for a while and it was relying on the MDEV-21187
buggy behavior to detect a slow query.
Reviewer: Monty
(Patch from Monty, slightly amended)
Fix rowid filtering optimization in best_access_path():
== Ref access + rowid filtering ==
The cost computations compare #records and index-only scan cost
(keyread_tmp) to find out the per-record advantage one will get if
they skip reading full table record.
The computations produce wrong result when:
- the #records are "clipped down" with s->worst_seeks or
thd->variables.max_seeks_for_key. keyread_tmp is not clipped
this way so the numbers are not comparable.
- access_factor is negative. This means index_only read is
cheaper than non-index-only read.
This patch makes the optimizer not to consider Rowid Filtering in
such cases.
The decision is logged in the Optimizer Trace using
"rowid_filter_skipped" name.
== Range access + rowid filtering ==
when considering to use Rowid Filter with range access, do multiply
keyread_tmp by record_count. That way, it is comparable with the
range access's estimate, which is multiplied by record_count.
This commit fixes the test system hanging due to
the galera_var_notify_ssl_ipv6 test and also brings
the wsrep_notify[_ssl].sh files in line with each other
between the user template and the mtr suite.
Quotes are also added here to avoid problems if the
user specifies the value of one of the variables at the
beginning of the file containing shell-specific characters,
for example, if the password or username specified in the
PSWD and USER variables will contain the "$" character.
Also fixed an issue with automatic --ssl-verify-server-cert
option substitution when the corresponding value is set
by the user to "1" or "on".
Also fixed some tests here to avoid joining one of the nodes
to another cluster when the nodes are restarted from the mtr
side, which can lead to random failures when testing with
buildbot.
This commit fixes the test system hanging due to
the galera_var_notify_ssl_ipv6 test and also brings
the wsrep_notify[_ssl].sh files in line with each other
between the user template and the mtr suite.
Quotes are also added here to avoid problems if the
user specifies the value of one of the variables at the
beginning of the file containing shell-specific characters,
for example, if the password or username specified in the
PSWD and USER variables will contain the "$" character.
Also fixed an issue with automatic --ssl-verify-server-cert
option substitution when the corresponding value is set
by the user to "1" or "on".
Also fixed some tests here to avoid joining one of the nodes
to another cluster when the nodes are restarted from the mtr
side, which can lead to random failures when testing with
buildbot.
mysql_discard_or_import_tablespace(): On successful
ALTER TABLE...DISCARD TABLESPACE, evict the table handle from the
table definition cache, so that ha_innobase::close() will be invoked,
like InnoDB expects to be the case. This will avoid an assertion failure
ut_a(table->get_ref_count() == 0) during IMPORT TABLESPACE.
ha_innobase::open(): Do not issue any ER_TABLESPACE_DISCARDED warning.
Member functions for DML will do that.
ha_innobase::truncate(), ha_innobase::check_if_supported_inplace_alter():
Issue ER_TABLESPACE_DISCARDED warnings, to compensate for the removal of
the warning in ha_innobase::open().
row_quiesce_write_indexes(): Only write information about committed
indexes. The ALTER TABLE t NOWAIT ADD INDEX(c) in the nondeterministic
test case will most of the time fail due to a metadata lock (MDL) timeout
and leave behind an uncommitted index.
Reviewed by: Sergei Golubchik
