If we have a 2+ node cluster which is replicating from an async master
and the binlog_format is set to STATEMENT and multi-row inserts are executed
on a table with an auto_increment column such that values are automatically
generated by MySQL, then the server node generates wrong auto_increment
values, which are different from what was generated on the async master.
In the title of the MDEV-9519 it was proposed to ban start slave on a Galera
if master binlog_format = statement and wsrep_auto_increment_control = 1,
but the problem can be solved without such a restriction.
The causes and fixes:
1. We need to improve processing of changing the auto-increment values
after changing the cluster size.
2. If wsrep auto_increment_control switched on during operation of
the node, then we should immediately update the auto_increment_increment
and auto_increment_offset global variables, without waiting of the next
invocation of the wsrep_view_handler_cb() callback. In the current version
these variables retain its initial values if wsrep_auto_increment_control
is switched on during operation of the node, which leads to inconsistent
results on the different nodes in some scenarios.
3. If wsrep auto_increment_control switched off during operation of the node,
then we must return the original values of the auto_increment_increment and
auto_increment_offset global variables, as the user has set. To make this
possible, we need to add a "shadow copies" of these variables (which stores
the latest values set by the user).
https://jira.mariadb.org/browse/MDEV-9519
with UNION ALL after INTERSECT
EXPLAIN EXTENDED erroneously showed UNION instead of UNION ALL in
the warning if UNION ALL followed INTERSECT or EXCEPT operations.
The bug was in the function st_select_lex_unit::print() that printed
the text of the query used in the warning.
- CREATE TABLE ... SELECT drops constraints for columns that
are both in the create and select part.
- Fixed by copying the constraint in
Column_definiton::redefine_stage1_common()
- If one has both a default expression and check constraint for a
column, one can get the error "Expression for field `a` is refering
to uninitialized field `a`.
- Fixed by ignoring default expressions for current column when checking
for CHECK constraint
No need to call list.empty(): first one is called by List constructor,
second one doesn't make sense as the object is destroyed immediately
afterwards.
renaming columns in a CHECK constraint during ALTER TABLE
taints the original TABLE and requires m_need_reopen=1.
In this case, though, renaming was redundant, so just don't do it.
remove TABLE_SHARE::error_table_name() and TABLE_SHARE::orig_table_name
(that was allocated in a wrong memroot in this bug).
instead, simply set TABLE_SHARE::table_name correctly.
wsrep_certification_rules: Define as a weak global symbol.
While there are separate _embedded.a for statically
linked storage engine plugins, there is only one ha_innodb.so
which is supposed to work with both values of WITH_WSREP.
The merge from 10.0-galera introduced a reference to a global
variable that is only defined when the server is built WITH_WSREP.
We must define that symbol as weak global, so that when
a dynamically linked InnoDB or XtraDB is used with the embedded
server (which never includes write-set replication patches),
the variable will be read as 0, instead of causing a failure to
load the InnoDB or XtraDB plugin.
No need to lowercase table names on case-sensitive file systems, as the
cache won't contain the 'lowercased' table anyway. And it prevents the
UPPERCASE.frm from being deleted.
Do not try to write ER_SHUTDOWN error message to socket, when it is forcefully closed by the shutdown.
This will avoid the race condition (attempt to write to closed socket, if connection shuts down by itself).
Analysis:
========
Increasing the length of the indexed varchar column is not an instant operation for
innodb.
Fix:
===
- Introduce the new handler flag 'Alter_inplace_info::ALTER_COLUMN_INDEX_LENGTH' to
indicate the index length differs due to change of column length changes.
- InnoDB makes the ALTER_COLUMN_INDEX_LENGTH flag as instant operation.
This is a port of Mysql fix.
commit 913071c0b16cc03e703308250d795bc381627e37
Author: Nisha Gopalakrishnan <nisha.gopalakrishnan@oracle.com>
Date: Wed May 30 14:54:46 2018 +0530
BUG#26848813: INDEXED COLUMN CAN'T BE CHANGED FROM VARCHAR(15)
TO VARCHAR(40) INSTANTANEOUSLY
If the TC log did not provide list of XIDs to recover, the
commit by XID was skipped during wsrep recovery if binlog emulation
was on. However, with wsrep we want to commit every prepared transaction
with assigned wsrep XID since the transaction has already been
committed in the cluster.
Added a special condition to always proceed to commit by XID in
xarecover_handlerton() if binlog is off and the recovered transaction
has wsrep XID.
This patch contains the port of the MDEV-18379 patch
for 10.1 branch, but also includes a number of changes
made within MDEV-17835, which are necessary for the
normal operation of tests that use IPv6:
1) Fixed flaws in the galera_3nodes mtr suite control scripts,
because of which they could not work with mariabackup.
2) Fixed numerous bugs in the SST scripts and in the mtr test
files (galera_3nodes mtr suite) that prevented the use of Galera
with IPv6 addresses.
3) Fixed flaws in tests for rsync and mysqldump (for galera_3nodes
mtr tests suite). These tests were not performed successfully
without these fixes.
4) Currently, the three-node mtr suite for Galera (galera_3nodes)
uses a separate IPv6 availability check using the "have_ipv6.inc"
file. This check duplicates a more accurate check at suite.pm
level, which can be used by including the file "check_ipv6.inc".
This patch removes this discrepancy between suites.
5) GAL-501 test in the galera_3nodes suite does not contain the
option "--bind-address=::" which is needed for the test to work
correctly with IPv6 (at least on some systems), since without
it the server will not wait for connections on the IPv6 interface.
https://jira.mariadb.org/browse/MDEV-18379
and partially https://jira.mariadb.org/browse/MDEV-17835
Clear wsrep XID in innobase_rollback_by_xid() for recovered wsrep
transaction in order to avoid resetting XID storage when rolling back
wsrep transaction during recovery.
Sort wsrep XIDs read from storage engine in ascending order and
erify that the range is continuous during crash recovery. If binlog is off,
commit all recovered transactions for continuous seqno range. This is safe
because all transactions with wsrep XID have been certified and must be
committed in the cluster. On the other hand if binlog is on, respect binlog
as a transaction coordinator in order to avoid missing transactions in binlog
that have been committed into storage engine .
The problem was originally stated in
http://bugs.mysql.com/bug.php?id=82212
The size of an base64-encoded Rows_log_event exceeds its
vanilla byte representation in 4/3 times.
When a binlogged event size is about 1GB mysqlbinlog generates
a BINLOG query that can't be send out due to its size.
It is fixed with fragmenting the BINLOG argument C-string into
(approximate) halves when the base64 encoded event is over 1GB size.
The mysqlbinlog in such case puts out
SET @binlog_fragment_0='base64-encoded-fragment_0';
SET @binlog_fragment_1='base64-encoded-fragment_1';
BINLOG @binlog_fragment_0, @binlog_fragment_1;
to represent a big BINLOG.
For prompt memory release BINLOG handler is made to reset the BINLOG argument
user variables in the middle of processing, as if @binlog_fragment_{0,1} = NULL
is assigned.
Notice the 2 fragments are enough, though the client and server still may
need to tweak their @@max_allowed_packet to satisfy to the fragment
size (which they would have to do anyway with greater number of
fragments, should that be desired).
On the lower level the following changes are made:
Log_event::print_base64()
remains to call encoder and store the encoded data into a cache but
now *without* doing any formatting. The latter is left for time
when the cache is copied to an output file (e.g mysqlbinlog output).
No formatting behavior is also reflected by the change in the meaning
of the last argument which specifies whether to cache the encoded data.
Rows_log_event::print_helper()
is made to invoke a specialized fragmented cache-to-file copying function
which is
copy_cache_to_file_wrapped()
that takes care of fragmenting also optionally wraps encoded
strings (fragments) into SQL stanzas.
my_b_copy_to_file()
is refactored to into my_b_copy_all_to_file(). The former function
is generalized
to accepts more a limit argument to constraint the copying and does
not reinitialize anymore the cache into reading mode.
The limit does not do any effect on the fully read cache.
always logged properly with binlog_row_image=MINIMAL
There are two issues fixed in this commit.
The first is an observation of a multi-table UPDATE binlogged
in row-format in binlog_row_image=MINIMAL mode. While the UPDATE aims
at a table with an ON-UPDATE attribute its binlog after-image misses
to record also installed default value.
The reason for that turns out missed marking of default-capable fields
in TABLE::write_set.
This is fixed to mark such fields similarly to 10.2's MDEV-10134 patch (db7edfed17)
that introduced it. The marking follows up 93d1e5ce0b841bed's idea
to exploit TABLE:rpl_write_set introduced there though,
and thus does not mess (in 10.1) with the actual MDEV-10134 agenda.
The patch makes formerly arg-less TABLE::mark_default_fields_for_write()
to accept an argument which would be TABLE:rpl_write_set.
The 2nd issue is extra columns in in binlog_row_image=MINIMAL before-image
while merely a packed primary key is enough. The test main.mysqlbinlog_row_minimal
always had a wrong result recorded.
This is fixed to invoke a function that intended for read_set
possible filtering and which is called (supposed to) in all type of MDL, UPDATE
including; the test results have gotten corrected.
At *merging* from 10.1->10.2 the 1st "main" part of the patch is unnecessary
since the bug is not observed in 10.2, so only hunks from
sql/sql_class.cc
are required.
Calling st_select_lex::update_used_tables in JOIN::optimize_unflattened_subqueries
only when we are sure that the join have not been cleaned up.
This can happen for a case when we have a non-merged semi-join and an impossible
where which would lead to the cleanup of the join which has the non-merged semi-join
ASAN noticed a freed memory access during EXECUTE in this script:
PREPARE stmt FROM "SELECT 'x' ORDER BY NAME_CONST( 'f', 'foo' )";
EXECUTE stmt;
In case of a PREPARE statement, all Items, including Item_name_const,
are created on Prepared_statement::main_mem_root.
Item_name_const::fix_fields() did not take this into account
and could allocate the value of Item::name on a wrong memory root,
in this code:
if (is_autogenerated_name)
{
set_name(thd, item_name->c_ptr(), (uint) item_name->length(),
system_charset_info);
}
When fix_fields() is called in the reported SQL script, THD's arena already
points to THD::main_mem_root rather than to Prepared_statement::main_mem_root,
so Item::name was allocated on THD::main_mem_root.
Then, at the end of the dispatch_command() for the PREPARE statement,
THD::main_mem_root got cleared. So during EXECUTE, Item::name
pointed to an already freed memory.
This patch changes the code to set the implicit name for Item_name_const
at the constructor time rather than at fix_fields time. This guarantees
that Item_name_const and its Item::name always reside on the same memory root.
Note, this change makes the code for Item_name_const symmetric with other
constant-alike items that set their default implicit names at the constructor
call time rather than at fix_fields() time:
- Item_string
- Item_int
- Item_real
- Item_decimal
- Item_null
- Item_param
Problem:
========
Server fails to notify the engine by not setting the ADD_PK_INDEX and
DROP_PK_INDEX When there is a
i) Change in candidate for primary key.
ii) New candidate for primary key.
Fix:
====
Server sets the ADD_PK_INDEX and DROP_PK_INDEX while doing alter for the
above problematic case.
32 bit int
Row-based slave applier could not parse correctly the table id when
the value exceeded the max of 32 bit unsigned int.
The reason turns out in that the being parsed value placeholder
was sized as 4 bytes.
The type is fixed to ulonglong.
Additionally the patch works around Rows_log_event::m_table_id 4 bytes
size on 32 bits platforms. In case of last_table_id value overflows
the 4 byte max, there won't be the zero value for m_table_id generated
and the first wrapped-around value is one, this is thanks to excluding
UINT_MAX32 + 1 from TABLE_SHARE::table_map_id.
Issue:
------
When a subquery contains UNION the count of the number of
subquery columns is calculated incorrectly. Only the first
query block in the subquery's UNION is considered and an
array indexing goes out-of-bounds, and this is caught by an
assert.
Solution:
---------
Sum up the columns from all query blocks of the query
expression.
Change specific to 5.6/5.5:
---------------------------
The "child" points to the last query block of the UNION
(as opposed to 5.7+ where it points to the first member of
UNION). So "child->master_unit()->first_select()" is used
to reach the first query block of UNION.
