The current semi-sync binlog fail-over recovery process uses
rpl_semi_sync_slave_enabled==TRUE as its condition to truncate a
primary server’s binlog, as it is anticipating the server to re-join
a replication topology as a replica. However, for servers configured
with both rpl_semi_sync_master_enabled=1 and
rpl_semi_sync_slave_enabled=1, if a primary is just re-started (i.e.
retaining its role as master), it can truncate its binlog to drop
transactions which its replica(s) has already received and executed.
If this happens, when the replica reconnects, its gtid_slave_pos can
be ahead of the recovered primary’s gtid_binlog_pos, resulting in an
error state where the replica’s state is ahead of the primary’s.
This patch changes the condition for semi-sync recovery to truncate
the binlog to instead use the configuration variable
--init-rpl-role, when set to SLAVE. This allows for both
rpl_semi_sync_master_enabled and rpl_semi_sync_slave_enabled to be
set for a primary that is restarted, and no transactions will be
lost, so long as --init-rpl-role is not set to SLAVE.
Reviewed By:
============
Sergei Golubchik <serg@mariadb.com>
The special logic used by the memory storage engine
to keep slaves in sync with the master on a restart can
break replication. In particular, after a restart, the
master writes DELETE statements in the binlog for
each MEMORY-based table so the slave can empty its
data. If the DELETE is not executable, e.g. due to
invalid triggers, the slave will error and fail, whereas
the master will never see the problem.
Instead of DELETE statements, use TRUNCATE to
keep slaves in-sync with the master, thereby bypassing
triggers.
Reviewed By:
===========
Kristian Nielsen <knielsen@knielsen-hq.org>
Andrei Elkin <andrei.elkin@mariadb.com>
During read only mode, InnoDB doesn't allow checkpoint to happen.
So InnoDB should throw the warning when InnoDB tries to
force the checkpoint when innodb_read_only = 1 or
innodb_force_recovery = 6.
MariaDB supports a "wait-free concurrent allocator based on pinning addresses".
In `lf_pinbox_real_free()` it tries to sort the pinned addresses for better
performance to use binary search during "real free". `alloca()` was used to
allocate stack memory and copy addresses.
To prevent a stack overflow when allocating the stack memory the function checks
if there's enough stack space. However, the available stack size was calculated
inaccurately which eventually caused database crash due to stack overflow.
The crash was seen on MariaDB 10.6.11 but the same code defect exists on all
MariaDB versions.
A similar issue happened previously and the fix in fc2c1e43 was to add a
`ALLOCA_SAFETY_MARGIN` which is 8192 bytes. However, that safety margin is not
enough during high connection workloads.
MySQL also had a similar issue and the fix
https://github.com/mysql/mysql-server/commit/b086fda was to remove the use of
`alloca` and replace qsort approach by a linear scan through all pointers (pins)
owned by each thread.
This commit is mostly the same as it is the only way to solve this issue as:
1. Frame sizes in different architecture can be different.
2. Number of active (non-null) pinned addresses varies, so the frame
size for the recursive sorting function `msort_with_tmp` is also hard
to predict.
3. Allocating big memory blocks in stack doesn't seem to be a very good
practice.
For further details see the mentioned commit in MySQL and the inline comments.
All new code of the whole pull request, including one or several files
that are either new files or modified ones, are contributed under the
BSD-new license. I am contributing on behalf of my employer Amazon Web
Services, Inc.
The symptoms were: take a server with no activity and a table that's
not in the buffer pool. Run a query that reads the whole table and
observe that r_engine_stats.pages_read_count shows about 2% of the table
was read. Who reads the rest?
The cause was that page prefetching done inside InnoDB was not counted.
This counts page prefetch requests made in buf_read_ahead_random() and
buf_read_ahead_linear() and makes them visible in:
- ANALYZE: r_engine_stats.pages_prefetch_read_count
- Slow Query Log: Pages_prefetched:
This patch intentionally doesn't attempt to count the time to read the
prefetched pages:
* there's no obvious place where one can do it
* prefetch reads may be done in parallel (right?), it is not clear how
to count the time in this case.
The crash is caused by the attempt to refix the constant subquery during
pushdown from HAVING into WHERE optimization.
Every condition that is going to be pushed into WHERE clause is first
cleaned up, then refixed. Constant subqueries are not cleaned or refixed
because they will remain the same after refixing, so this complicated
procedure can be omitted for them (introduced in MDEV-21184).
Constant subqueries are marked with flag IMMUTABLE_FL, that helps to miss
the cleanup stage for them. Also they are marked as fixed, so refixing is
also not done for them.
Because of the multiple equality propagation several references to the same
constant subquery can exist in the condition that is going to be pushed
into WHERE. Before this patch, the problem appeared in the following way.
After the first reference to the constant subquery is processed, the flag
IMMUTABLE_FL for the constant subquery is disabled.
So, when the second reference to this constant subquery is processed, the
flag is already disabled and the subquery goes through the procedure of
cleaning and refixing. That causes a crash.
To solve this problem, IMMUTABLE_FL should be disabled only after all
references to the constant subquery are processed, so after the whole
condition that is going to be pushed is cleaned up and refixed.
Approved by Igor Babaev <igor@maridb.com>
Fixing applying the COLLATE clause to a parameter caused an error error:
COLLATION '...' is not valid for CHARACTER SET 'binary'
Fix:
- Changing the collation derivation for a non-prepared Item_param
to DERIVATION_IGNORABLE.
- Allowing to apply any COLLATE clause to expressions with DERIVATION_IGNORABLE.
This includes:
1. A non-prepared Item_param
2. An explicit NULL
3. Expressions derived from #1 and #2
For example:
SELECT ? COLLATE utf8mb_unicode_ci;
SELECT NULL COLLATE utf8mb_unicode_ci;
SELECT CONCAT(?) COLLATE utf8mb_unicode_ci;
SELECT CONCAT(NULL) COLLATE utf8mb_unicode_ci
- Additional change: preserving the collation of an expression when
the expression gets assigned to a PS parameter and evaluates to SQL NULL.
Before this change, the collation of the parameter was erroneously set
to &my_charset_binary.
- Additional change: removing the multiplication to mbmaxlen from the
fix_char_length_ulonglong() argument, because the multiplication already
happens inside fix_char_length_ulonglong().
This fixes a too large column size created for a COLLATE clause.
In 10.0 there was an assert to ensure that there were semi
sync clients before removing one, but it was removed in 10.1.
This patch adds the assertion back.
The performance regression seen while loading BP is caused by the
deadlock fix given in MDEV-33543. The area of impact is wider but is
more visible when BP is being loaded initially via DMLs. Specifically
the response time could be impacted in DML doing pessimistic operation
on index(split/merge) and the leaf pages are not found in buffer pool.
It is more likely to occur with small BP size.
The origin of the issue dates back to MDEV-30400 that introduced
btr_cur_t::search_leaf() replacing btr_cur_search_to_nth_level() for
leaf page searches. In btr_latch_prev, we use RW_NO_LATCH to get the
previous page fixed in BP without latching. When the page is not in BP,
we try to acquire and wait for S latch violating the latching order.
This deadlock was analyzed in MDEV-33543 and fixed by using the already
present wait logic in buf_page_get_gen() instead of waiting for latch.
The wait logic is inferior to usual S latch wait and is simply a
repeated sleep 100 of micro-sec (The actual sleep time could be more
depending on platforms). The bug was seen with "change-buffering" code
path and the idea was that this path should be less exercised. The
judgement was not correct and the path is actually quite frequent and
does impact performance when pages are not in BP and being loaded by
DML expanding/shrinking large data.
Fix: While trying to get a page with RW_NO_LATCH and we are attempting
"out of order" latch, return from buf_page_get_gen immediately instead
of waiting and follow the ordered latching path.
Valgrind looks as the assertions as examining uninitalized values.
As the assertions are tested in other Debug builds we know
it isn't all invalid.
Account for Valgrind by removing the assertion under
the WITH_VALGRIND=1 compulation.
The memory leak happened on second execution of a prepared statement
that runs UPDATE statement with correlated subquery in right hand side of
the SET clause. In this case, invocation of the method
table->stat_records()
could return the zero value that results in going into the 'if' branch
that handles impossible where condition. The issue is that this condition
branch missed saving of leaf tables that has to be performed as first
condition optimization activity. Later the PS statement memory root
is marked as read only on finishing first time execution of the prepared
statement. Next time the same statement is executed it hits the assertion
on attempt to allocate a memory on the PS memory root marked as read only.
This memory allocation takes place by the sequence of the following
invocations:
Prepared_statement::execute
mysql_execute_command
Sql_cmd_dml::execute
Sql_cmd_update::execute_inner
Sql_cmd_update::update_single_table
st_select_lex::save_leaf_tables
List<TABLE_LIST>::push_back
To fix the issue, add the flag SELECT_LEX::leaf_tables_saved to control
whether the method SELECT_LEX::save_leaf_tables() has to be called or
it has been already invoked and no more invocation required.
Similar issue could take place on running the DELETE statement with
the LIMIT clause in PS/SP mode. The reason of memory leak is the same as for
UPDATE case and be fixed in the same way.
Caused by:
5d37cac7 MDEV-33348 ALTER TABLE lock waiting stages are indistinguishable.
In that commit, progress reporting was moved to
mysql_alter_table from copy_data_between_tables.
The temporary table case wasn't taken into the consideration,
where the execution of mysql_alter_table ends earlier than usual, by the
'end_temporary' label. There, thd_progress_end has been missing.
Fix:
Add missing thd_progress_end() call in mysql_alter_table.
The feedback plugin server_uid variable and the calculate_server_uid()
function is moved from feedback/utils.cc to sql/mysqld.cc
server_uid is added as a global variable (shown in 'show variables') and
is written to the error log on server startup together with server version
and server commit id.
We have an issue if a user have the following in a configuration file:
log_slow_filter="" # Log everything to slow query log
log_queries_not_using_indexes=ON
This set log_slow_filter to 'not_using_index' which disables
slow_query_logging of most queries.
In effect, on should never use log_slow_filter="" in config files but
instead use log_slow_filter=ALL.
Fixed by changing log_slow_filter="" that comes either from a
configuration file or from the command line, when starting to the server,
to log_slow_filter=ALL.
A warning will be printed when this happens.
Other things:
- One can now use =ALL for any 'set' variable to set all options at once.
(backported from 10.6)
When getaddrinfo returns and error, the contents
of ai are invalid so we cannot continue based
on their data structures.
In the previous branch of the if statement, we
abort there if there is an error so for consistency
we abort here too.
The test case fixes the port number to UINTMAX32
for both an enumberated bind-address and the
default bind-address covering the two calls to
getaddrinfo.
Review thanks Sanja.
Item_func_hex::fix_length_and_dec() evaluated a too short data type
for signed numeric arguments, which resulted in a 'Data too long for column'
error on CREATE..SELECT.
Fixing the code to take into account that a short negative
numer can produce a long HEX value: -1 -> 'FFFFFFFFFFFFFFFF'
Also fixing Item_func_hex::val_str_ascii_from_val_real().
Without this change, MTR test with HEX with negative float point arguments
failed on some platforms (aarch64, ppc64le, s390-x).
InnoDB transactions may be reused after committed:
- when taken from the transaction pool
- during a DDL operation execution
In this case wsrep flag on trx object is cleared, which may cause wrong
execution logic afterwards (wsrep-related hooks are not run).
Make trx->wsrep flag initialize from THD object only once on InnoDB transaction
start and don't change it throughout the transaction's lifetime.
The flag is reset at commit time as before.
Unconditionally set wsrep=OFF for THD objects that represent InnoDB background
threads.
Make Wsrep_schema::store_view() operate in its own transaction.
Fix streaming replication transactions' fragments rollback to not switch
THD->wsrep value during transaction's execution
(use THD->wsrep_ignore_table as a workaround).
Signed-off-by: Julius Goryavsky <julius.goryavsky@mariadb.com>
Affects:
MDEV-34150 Assertion failure in Diagnostics_area::set_error_status upon binary
logging hitting tmp space limit
MDEV-9101 Limit size of created disk temporary files and tables
This bug was caused by moving flushing of the in-memory-row-events from
close_thread_tables() to binlog_commit() in MDEV-34150.
This was needed to be able to handle the case where binlog writes could
fail.
Galera have two case where the change caused problems:
- Row events in commit_one_phase_2() was not done in the case the standard
binary log was not enabled but Galera was using the binary log
internally.
- Galera disabled the call to binlog_commit_flush_stmt_cache() for not
ending transactions.
Fixed by adding code that flushes the in-memory-row-events to the binary
log (write, but now sync) in the two above cases if Galera is enabled.
realloc can't be done for mysql->passwd. There's also no pretty way to
pass the ext-salt into hash_password, so let's use thread_local instead.
For GCC < 4.9 (SLES 12, CentOS 7), which don't support _Thread_local,
pk will be recalculated instead
During a Sysbench oltp_point_select workload with 1 table and 400
concurrent connections, a bottleneck on dict_table_t::lock_mutex was
observed in ha_innobase::info_low().
dict_table_t::lock_latch: Replaces lock_mutex.
In ha_innobase::info_low() and several other places, we will acquire
a shared dict_table_t::lock_latch or we may elide the latch if
hardware memory transactions are available.
innobase_build_v_templ(): Remove the parameter "bool locked", and
require the caller to hold exclusive dict_table_t::lock_latch
(instead of holding an exclusive dict_sys.latch).
Tested by: Vladislav Vaintroub
Reviewed by: Vladislav Vaintroub
As part of commit 685d958e38 (MDEV-14425)
the parameter innodb_log_write_ahead_size was removed, because it was
thought that determining the physical block size would be a sufficient
replacement.
However, we can only determine the physical block size on Linux or
Microsoft Windows. On some file systems, the physical block size
is not relevant. For example, XFS uses a block size of 4096 bytes
even if the underlying block size may be smaller.
On Linux, we failed to determine the physical block size if
innodb_log_file_buffered=OFF was not requested or possible.
This will be fixed.
log_sys.write_size: The value of the reintroduced parameter
innodb_log_write_ahead_size. To keep it simple, this is read-only
and a power of two between 512 and 4096 bytes, so that the previous
alignment guarantees are fulfilled. This will replace the previous
log_sys.get_block_size().
log_sys.block_size, log_t::get_block_size(): Remove.
log_t::set_block_size(): Ensure that write_size will not be less
than the physical block size. There is no point to invoke this
function with 512 or less, because that is the minimum value of
write_size.
innodb_params_adjust(): Add some disabled code for adjusting
the minimum value and default value of innodb_log_write_ahead_size
to reflect the log_sys.write_size.
log_t::set_recovered(): Mark the recovery completed. This is the
place to adjust some things if we want to allow write_size>4096.
log_t::resize_write_buf(): Refer to write_size.
log_t::resize_start(): Refer to write_size instead of get_block_size().
log_write_buf(): Simplify some arithmetics and remove a goto.
log_t::write_buf(): Refer to write_size. If we are writing less than
that, do not switch buffers, but keep writing to the same buffer.
Move some code to improve the locality of reference.
recv_scan_log(): Refer to write_size instead of get_block_size().
os_file_create_func(): For type==OS_LOG_FILE on Linux, always invoke
os_file_log_maybe_unbuffered(), so that log_sys.set_block_size() will
be invoked even if we are not attempting to use O_DIRECT.
recv_sys_t::find_checkpoint(): Read the entire log header
in a single 12 KiB request into log_sys.buf.
Tested with:
./mtr --loose-innodb-log-write-ahead-size=4096
./mtr --loose-innodb-log-write-ahead-size=2048
PFS_atomic class contains wrappers around my_atomic_* operations, which
are macros to GNU atomic operations (__atomic_*). Due to different
implementations of compilers, clang may encounter errors when compiling
on x86_32 architecture.
The following functions are replaced with C++ std::atomic type in
performance schema code base:
- PFS_atomic::store_*()
-> my_atomic_store*
-> __atomic_store_n()
=> std::atomic<T>::store()
- PFS_atomic::load_*()
-> my_atomic_load*
-> __atomic_load_n()
=> std::atomic<T>::load()
- PFS_atomic::add_*()
-> my_atomic_add*
-> __atomic_fetch_add()
=> std::atomic<T>::fetch_add()
- PFS_atomic::cas_*()
-> my_atomic_cas*
-> __atomic_compare_exchange_n()
=> std::atomic<T>::compare_exchange_strong()
and PFS_atomic class could be dropped completely.
Note that in the wrapper memory order passed to original GNU atomic
extensions are hard-coded as `__ATOMIC_SEQ_CST`, which is equivalent to
`std::memory_order_seq_cst` in C++, and is the default parameter for
std::atomic_* functions.
All new code of the whole pull request, including one or several files
that are either new files or modified ones, are contributed under the
BSD-new license. I am contributing on behalf of my employer Amazon Web
Services.
BTR_MODIFY_PREV: Remove. This mode was only used by the change buffer,
which commit f27e9c8947 (MDEV-29694)
removed.
buf_page_get_gen(): Revert the change that was made in
commit 90b95c6149 (MDEV-33543)
because it is not applicable after MDEV-29694. This fixes the
performance regression that Vladislav Vaintroub reported.
This is a 11.x specific fix; this needs to be fixed differently
in older major versions where the change buffer is present.
innodb_doublewrite_update(): Disallow any change if srv_read_only_mode
holds, that is, the server was started with innodb_read_only=ON or
innodb_force_recovery=6.
This fixes up commit 1122ac978e (MDEV-33545).
CMake WSREP=ON has some implications for client
executables so still present this as an option
when compiling WITHOUT_SERVER. In this case
default to ON for maximium compatibility of
the build client executables and libraries.
The log_event_old.cc is included by mysqlbinlog.cc.
With -DWITHOUT_SERVER the include path for the wsrep
include headers isn't there.
As these aren't needed by the mariadb-binlog, move
these to under a ifndef MYSQL_CLIENT preprocessor.
Caused by MDEV-18590
Just like the spider/bugfix suite.
One caveat is that my_2_3.cnf needs something under mysqld.2.3 group,
otherwise mtr will fail with something like:
There is no group named 'mysqld.2.3' that can be used to resolve
'port' for ...
This will allow new tests under the spider suite to use what is
needed. It also somehow fixes issues of running a test followed by
spider.slave_trx_isolation.
One of possible use cases that reproduces the memory leakage listed below:
set timestamp= unix_timestamp('2000-01-01 00:00:00');
create or replace table t1 (x int) with system versioning
partition by system_time interval 1 hour auto
partitions 3;
create table t2 (x int);
create trigger tr after insert on t2 for each row update t1 set x= 11;
create or replace procedure sp2() insert into t2 values (5);
set timestamp= unix_timestamp('2000-01-01 04:00:00');
call sp2;
set timestamp= unix_timestamp('2000-01-01 13:00:00');
call sp2; # <<=== Memory leak happens there. In case MariaDB server is built
with the option -DWITH_PROTECT_STATEMENT_MEMROOT,
the second execution would hit assert failure.
The reason of leaking a memory is that once a new partition be created
the table should be closed and re-opened. It results in calling the function
extend_table_list() that indirectly invokes the function sp_add_used_routine()
to add routines implicitly used by the statement that makes a new memory
allocation.
To fix it, don't remove routines and tables the statement implicitly depends
on when a table being closed for subsequent re-opening.
Executing an INSERT statement in PS mode having positional parameter
bound with an array could result in incorrect number of inserted rows
in case there is a BEFORE INSERT trigger that executes yet another
INSERT statement to put a copy of row being inserted into some table.
The reason for incorrect number of inserted rows is that a data structure
used for binding positional argument with its actual values is stored
in THD (this is thd->bulk_param) and reused on processing every INSERT
statement. It leads to consuming actual values bound with top-level
INSERT statement by other INSERT statements used by triggers' body.
To fix the issue, reset the thd->bulk_param temporary to the value nullptr
before invoking triggers and restore its value on finishing its execution.
