backport for bug#44059 from mysql-pe to mysql-5.1-bugteam
Using the partition with most rows instead of first partition
to estimate the cardinality of indexes.
is reached
Problem was bad error handling, leaving some new temporary
partitions locked and initialized and some not yet initialized
and locked, leading to a crash when trying to unlock the not
yet initialized and locked partitions
Solution was to unlock the already locked partitions, and not
include any of the new temporary partitions in later unlocks
- Read plug.in to fid the name of the engine to link with, does not have
to be same as engine dir
- Use engine dir when figuring out which libraries to build limbysqld with
From revisions r5792 to r5864
Detailed revision comments:
r5792 | vasil | 2009-09-09 08:35:58 -0500 (Wed, 09 Sep 2009) | 32 lines
branches/zip:
Fix a bug in manipulating the variable innodb_old_blocks_pct:
for any value assigned it got that value -1, except for 75. When
assigned 75, it got 75.
mysql> set global innodb_old_blocks_pct=15;
Query OK, 0 rows affected (0.00 sec)
mysql> show variables like 'innodb_old_blocks_pct';
+-----------------------+-------+
| Variable_name | Value |
+-----------------------+-------+
| innodb_old_blocks_pct | 14 |
+-----------------------+-------+
1 row in set (0.00 sec)
mysql> set global innodb_old_blocks_pct=75;
Query OK, 0 rows affected (0.00 sec)
mysql> show variables like 'innodb_old_blocks_pct';
+-----------------------+-------+
| Variable_name | Value |
+-----------------------+-------+
| innodb_old_blocks_pct | 75 |
+-----------------------+-------+
After the fix it gets exactly what was assigned.
Approved by: Marko (via IM)
r5798 | calvin | 2009-09-09 10:28:10 -0500 (Wed, 09 Sep 2009) | 5 lines
branches/zip:
HA_ERR_TOO_MANY_CONCURRENT_TRXS is added in 5.1.38.
But the plugin should still work with previous versions
of MySQL.
r5804 | marko | 2009-09-10 00:29:31 -0500 (Thu, 10 Sep 2009) | 1 line
branches/zip: trx_cleanup_at_db_startup(): Fix a typo in comment.
r5822 | marko | 2009-09-10 05:10:20 -0500 (Thu, 10 Sep 2009) | 1 line
branches/zip: buf_page_release(): De-stutter the function comment.
r5825 | marko | 2009-09-10 05:47:09 -0500 (Thu, 10 Sep 2009) | 20 lines
branches/zip: Reduce mutex contention that was introduced when
addressing Bug #45015 (Issue #316), in r5703.
buf_page_set_accessed_make_young(): New auxiliary function, called by
buf_page_get_zip(), buf_page_get_gen(),
buf_page_optimistic_get_func(). Call ut_time_ms() outside of
buf_pool_mutex. Use cached access_time.
buf_page_set_accessed(): Add the parameter time_ms, so that
ut_time_ms() need not be called while holding buf_pool_mutex.
buf_page_optimistic_get_func(), buf_page_get_known_nowait(): Read
buf_page_t::access_time without holding buf_pool_mutex. This should be
OK, because the field is only used for heuristic purposes.
buf_page_peek_if_too_old(): If buf_pool->freed_page_clock == 0, return
FALSE, so that we will not waste time moving blocks in the LRU list in
the warm-up phase or when the workload fits in the buffer pool.
rb://156 approved by Sunny Bains
r5826 | marko | 2009-09-10 06:29:46 -0500 (Thu, 10 Sep 2009) | 12 lines
branches/zip: Roll back recovered dictionary transactions before
dropping incomplete indexes (Issue #337).
trx_rollback_or_clean_recovered(ibool all): New function, split from
trx_rollback_or_clean_all_recovered(). all==FALSE will only roll back
dictionary transactions.
recv_recovery_from_checkpoint_finish(): Call
trx_rollback_or_clean_recovered(FALSE) before
row_merge_drop_temp_indexes().
rb://158 approved by Sunny Bains
r5858 | vasil | 2009-09-11 12:46:47 -0500 (Fri, 11 Sep 2009) | 4 lines
branches/zip:
Fix the indentation of the closing bracket.
r5863 | vasil | 2009-09-12 02:07:08 -0500 (Sat, 12 Sep 2009) | 10 lines
branches/zip:
Check that pthread_t can indeed be passed to Solaris atomic functions, instead
of assuming that it can be passed if 0 can be assigned to it. It could be that:
* 0 can be assigned, but pthread_t cannot be passed and
* 0 cannot be assigned but pthread_t can be passed
Better to check what we are interested in, not something else and make
assumptions.
r5864 | vasil | 2009-09-12 02:22:55 -0500 (Sat, 12 Sep 2009) | 4 lines
branches/zip:
Include string.h which is needed for memset().
- Make it possible for the CmakeLists.txt files in an engine to use
${engine}_LIBS to set additional libraries to link with
Example: NDBCLUSTER_LIBS = ndbclient
From revision r5748 to r5783
Detailed revision comments:
r5748 | marko | 2009-09-03 06:05:44 -0500 (Thu, 03 Sep 2009) | 1 line
branches/zip: MLOG_MULTI_REC_END: Correct the comment.
r5751 | marko | 2009-09-03 09:36:15 -0500 (Thu, 03 Sep 2009) | 7 lines
branches/zip: row_merge(): Remove a bogus debug assertion
that was triggered when creating an index on an empty table.
row_merge_sort(): Add debug assertions and comments that justify
the loop termination condition.
The bogus assertion ut_ad(ihalf > 0) was reported by Michael.
r5752 | marko | 2009-09-03 09:55:51 -0500 (Thu, 03 Sep 2009) | 10 lines
branches/zip: recv_recover_page_func(): Write the log sequence number
to the compressed page, if there is one. Previously, the function only
wrote the LSN to the uncompressed page.
It is not clear why recv_recover_page_func() is updating FIL_PAGE_LSN
in the buffer pool. The log sequence number will be stamped on the
page when it is flushed to disk, in buf_flush_init_for_writing().
I noticed this inconsistency when analyzing Issue #313, but this patch
does not fix it. That is no surprise, since FIL_PAGE_LSN should only
matter on disk files, not in the buffer pool.
r5775 | calvin | 2009-09-07 16:15:05 -0500 (Mon, 07 Sep 2009) | 13 lines
branches/zip: Build InnoDB on Windows with UNIV_HOTBACKUP
The changes are non-functional changes for normal InnoDB,
but needed for building the Hot Backup on Windows (with
UNIV_HOTBACKUP defined).
- Define os_aio_use_native_aio for HB.
- Do not acquire seek mutexes for backup since HB is single threaded.
- Do not use srv_flush_log_at_trx_commit for HB build
rb://155
Approved by: Marko
r5777 | marko | 2009-09-08 10:50:25 -0500 (Tue, 08 Sep 2009) | 2 lines
branches/zip: Remove BUF_LRU_INITIAL_RATIO, which should have been removed
together with buf_LRU_get_recent_limit().
r5779 | marko | 2009-09-09 01:17:19 -0500 (Wed, 09 Sep 2009) | 2 lines
branches/zip: buf_page_peek_if_too_old(): Make the bitmasking work when
buf_pool->freed_page_clock is wider than 32 bits.
r5780 | marko | 2009-09-09 01:50:50 -0500 (Wed, 09 Sep 2009) | 1 line
branches/zip: ut_time_ms(): Return ulint, not uint.
r5782 | marko | 2009-09-09 02:00:59 -0500 (Wed, 09 Sep 2009) | 2 lines
branches/zip: buf_page_peek_if_too_old(): Silence a compiler warning
that was introduced in r5779 on 32-bit systems.
r5783 | marko | 2009-09-09 02:25:00 -0500 (Wed, 09 Sep 2009) | 1 line
branches/zip: buf_page_is_accessed(): Correct the function comment.
From revision r5733 to r5747
Detailed revision comments:
r5733 | sunny | 2009-09-02 02:05:15 -0500 (Wed, 02 Sep 2009) | 6 lines
branches/zip: Fix a regression introduced by the fix for bug#26316. We check
whether a transaction holds any AUTOINC locks before we acquire the kernel
mutex and release those locks.
Fix for rb://153. Approved by Marko.
r5734 | sunny | 2009-09-02 02:08:45 -0500 (Wed, 02 Sep 2009) | 2 lines
branches/zip: Update ChangeLog with r5733 changes.
r5735 | marko | 2009-09-02 02:43:09 -0500 (Wed, 02 Sep 2009) | 2 lines
branches/zip: univ.i: Do not undefine PACKAGE or VERSION.
InnoDB source code does not refer to these macros.
r5736 | marko | 2009-09-02 02:53:19 -0500 (Wed, 02 Sep 2009) | 1 line
branches/zip: Enclose some timestamp functions in #ifndef UNIV_HOTBACKUP.
r5743 | marko | 2009-09-03 01:36:12 -0500 (Thu, 03 Sep 2009) | 3 lines
branches/zip: log_reserve_and_write_fast(): Remove the redundant
output parameter "success".
Success is also indicated by a nonzero return value.
r5744 | marko | 2009-09-03 03:28:35 -0500 (Thu, 03 Sep 2009) | 1 line
branches/zip: ut_align(): Make ptr const, like in ut_align_down().
r5745 | marko | 2009-09-03 03:38:22 -0500 (Thu, 03 Sep 2009) | 2 lines
branches/zip: log_check_log_recs(): Enclose in #ifdef UNIV_LOG_DEBUG.
Add const qualifiers.
r5746 | marko | 2009-09-03 03:55:36 -0500 (Thu, 03 Sep 2009) | 2 lines
branches/zip: log_reserve_and_write_fast(): Do not cache the log_sys pointer
in a local variable.
r5747 | marko | 2009-09-03 05:46:38 -0500 (Thu, 03 Sep 2009) | 2 lines
branches/zip: recv_scan_log_recs(): Replace while with do...while,
because the termination condition will always hold on the first iteration.
can lead to bad memory access
Problem: Field_bit is the only field which returns INT_RESULT
and doesn't have unsigned flag. As it's not a descendant of the
Field_num, so using ((Field_num *) field_bit)->unsigned_flag may lead
to unpredictable results.
Fix: check the field type before casting.
From revision r5703 to r5716
Detailed revision comments:
r5703 | marko | 2009-08-27 02:25:00 -0500 (Thu, 27 Aug 2009) | 41 lines
branches/zip: Replace the constant 3/8 ratio that controls the LRU_old
size with the settable global variable innodb_old_blocks_pct. The
minimum and maximum values are 5 and 95 per cent, respectively. The
default is 100*3/8, in line with the old behavior.
ut_time_ms(): New utility function, to return the current time in
milliseconds. TODO: Is there a more efficient timestamp function, such
as rdtsc divided by a power of two?
buf_LRU_old_threshold_ms: New variable, corresponding to
innodb_old_blocks_time. The value 0 is the default behaviour: no
timeout before making blocks 'new'.
bpage->accessed, bpage->LRU_position, buf_pool->ulint_clock: Remove.
bpage->access_time: New field, replacing bpage->accessed. Protected by
buf_pool_mutex instead of bpage->mutex. Updated when a page is created
or accessed the first time in the buffer pool.
buf_LRU_old_ratio, innobase_old_blocks_pct: New variables,
corresponding to innodb_old_blocks_pct
buf_LRU_old_ratio_update(), innobase_old_blocks_pct_update(): Update
functions for buf_LRU_old_ratio, innobase_old_blocks_pct.
buf_page_peek_if_too_old(): Compare ut_time_ms() to bpage->access_time
if buf_LRU_old_threshold_ms && bpage->old. Else observe
buf_LRU_old_ratio and bpage->freed_page_clock.
buf_pool_t: Add n_pages_made_young, n_pages_not_made_young,
n_pages_made_young_old, n_pages_not_made_young, for statistics.
buf_print(): Display buf_pool->n_pages_made_young,
buf_pool->n_pages_not_made_young. This function is only for crash
diagnostics.
buf_print_io(): Display buf_pool->LRU_old_len and quantities derived
from buf_pool->n_pages_made_young, buf_pool->n_pages_not_made_young.
This function is invoked by SHOW ENGINE INNODB STATUS.
rb://129 approved by Heikki Tuuri. This addresses Bug #45015.
r5704 | marko | 2009-08-27 03:31:17 -0500 (Thu, 27 Aug 2009) | 32 lines
branches/zip: Fix a critical bug in fast index creation that could
corrupt the created indexes.
row_merge(): Make "half" an in/out parameter. Determine the offset of
half the output file. Copy the last blocks record-by-record instead of
block-by-block, so that the records can be counted. Check that the
input and output have matching n_rec.
row_merge_sort(): Do not assume that two blocks of size N are merged
into a block of size 2*N. The output block can be shorter than the
input if the last page of each input block is almost empty. Use an
accurate termination condition, based on the "half" computed by
row_merge().
row_merge_read(), row_merge_write(), row_merge_blocks(): Add debug output.
merge_file_t, row_merge_file_create(): Add n_rec, the number of records
in the merge file.
row_merge_read_clustered_index(): Update n_rec.
row_merge_blocks(): Update and check n_rec.
row_merge_blocks_copy(): New function, for copying the last blocks in
row_merge(). Update and check n_rec.
This bug was discovered with a user-supplied test case that creates an
index where the initial temporary file is 249 one-megabyte blocks and
the merged files become smaller. In the test, possible merge record
sizes are 10, 18, and 26 bytes.
rb://150 approved by Sunny Bains. This addresses Issue #320.
r5705 | marko | 2009-08-27 06:56:24 -0500 (Thu, 27 Aug 2009) | 11 lines
branches/zip: dict_index_find_cols(): On column name lookup failure,
return DB_CORRUPTION (HA_ERR_CRASHED) instead of abnormally
terminating the server. Also, disable the previously added diagnostic
output to the error log, because mysql-test-run does not like extra
output in the error log. (Bug #44571)
dict_index_add_to_cache(): Handle errors from dict_index_find_cols().
mysql-test/innodb_bug44571.test: A test case for triggering the bug.
rb://135 approved by Sunny Bains.
r5706 | inaam | 2009-08-27 11:00:27 -0500 (Thu, 27 Aug 2009) | 20 lines
branches/zip rb://147
Done away with following two status variables:
innodb_buffer_pool_read_ahead_rnd
innodb_buffer_pool_read_ahead_seq
Introduced two new status variables:
innodb_buffer_pool_read_ahead = number of pages read as part of
readahead since server startup
innodb_buffer_pool_read_ahead_evicted = number of pages that are read
in as readahead but were evicted before ever being accessed since
server startup i.e.: a measure of how badly our readahead is
performing
SHOW INNODB STATUS will show two extra numbers in buffer pool section:
pages read ahead/sec and pages evicted without access/sec
Approved by: Marko
r5707 | inaam | 2009-08-27 11:20:35 -0500 (Thu, 27 Aug 2009) | 6 lines
branches/zip
Remove unused macros as we erased the random readahead code in r5703.
Also fixed some comments.
r5708 | inaam | 2009-08-27 17:43:32 -0500 (Thu, 27 Aug 2009) | 4 lines
branches/zip
Remove redundant TRUE : FALSE from the return statement
r5709 | inaam | 2009-08-28 01:22:46 -0500 (Fri, 28 Aug 2009) | 5 lines
branches/zip rb://152
Disable display of deprecated parameter innodb_file_io_threads in
'show variables'.
r5714 | marko | 2009-08-31 01:10:10 -0500 (Mon, 31 Aug 2009) | 5 lines
branches/zip: buf_chunk_not_freed(): Do not acquire block->mutex unless
block->page.state == BUF_BLOCK_FILE_PAGE. Check that block->page.state
makes sense.
Approved by Sunny Bains over the IM.
r5716 | vasil | 2009-08-31 02:47:49 -0500 (Mon, 31 Aug 2009) | 9 lines
branches/zip:
Fix Bug#46718 InnoDB plugin incompatible with gcc 4.1 (at least: on PPC): "Undefined symbol"
by implementing our own check in plug.in instead of using the result from
the check from MySQL because it is insufficient.
Approved by: Marko (rb://154)
From revision r5686 to r5702
r5696 is already applied
Detailed revision comments:
r5686 | vasil | 2009-08-20 04:15:05 -0500 (Thu, 20 Aug 2009) | 4 lines
branches/zip:
White-space fixup.
r5687 | vasil | 2009-08-20 04:20:22 -0500 (Thu, 20 Aug 2009) | 8 lines
branches/zip:
ChangeLog:
Follow the convention from the rest of the ChangeLog: for bugfixes from
bugs.mysql.com only the bug number and title goes in the ChangeLog. Detailed
explanation on what is the problem and how it was fixed is present in
the bugs database.
r5694 | marko | 2009-08-26 06:25:26 -0500 (Wed, 26 Aug 2009) | 2 lines
branches/zip: buf_page_t: Clarify that bpage->list may contain garbage.
This comment was provoked by Inaam.
r5695 | marko | 2009-08-26 08:14:59 -0500 (Wed, 26 Aug 2009) | 1 line
branches/zip: UNIV_DEBUG_LOCK_VALIDATE: Move the definition to univ.i.
r5697 | vasil | 2009-08-26 08:44:40 -0500 (Wed, 26 Aug 2009) | 4 lines
branches/zip:
Fix typo.
r5698 | inaam | 2009-08-26 09:34:35 -0500 (Wed, 26 Aug 2009) | 13 lines
branches/zip bug#42885 rb://148
The call to put IO threads to sleep was most probably meant for Windows
only as the comment in buf0rea.c suggests. However it was enabled on
all platforms. This patch restricts the sleep call to windows. This
approach of not putting threads to sleep makes even more sense because
now we have multiple threads working in the background and it probably
is not a good idea to put all of them to sleep because a user thread
wants to post a batch for readahead.
Approved by: Marko
r5701 | marko | 2009-08-27 02:01:42 -0500 (Thu, 27 Aug 2009) | 1 line
branches/zip: Document r5698 in the ChangeLog.
r5702 | marko | 2009-08-27 02:03:15 -0500 (Thu, 27 Aug 2009) | 1 line
branches/zip: Document also the files affected by r5698 in the ChangeLog.
From r5639 to r5685
Detailed revision comments:
r5639 | marko | 2009-08-06 05:39:34 -0500 (Thu, 06 Aug 2009) | 3 lines
branches/zip: mem_heap_block_free(): If innodb_use_sys_malloc is set,
do not tell Valgrind that the memory is free, to avoid
a bogus warning in Valgrind's built-in free() hook.
r5642 | calvin | 2009-08-06 18:04:03 -0500 (Thu, 06 Aug 2009) | 2 lines
branches/zip: remove duplicate "the" in comments.
r5662 | marko | 2009-08-11 04:54:16 -0500 (Tue, 11 Aug 2009) | 1 line
branches/zip: Bump the version number to 1.0.5 after releasing 1.0.4.
r5663 | marko | 2009-08-11 06:42:37 -0500 (Tue, 11 Aug 2009) | 2 lines
branches/zip: trx_general_rollback_for_mysql(): Remove the redundant
parameter partial. If savept==NULL, partial==FALSE.
r5670 | marko | 2009-08-12 08:16:37 -0500 (Wed, 12 Aug 2009) | 2 lines
branches/zip: trx_undo_rec_copy(): Add const qualifier to undo_rec.
This is a non-functional change.
r5671 | marko | 2009-08-13 03:46:33 -0500 (Thu, 13 Aug 2009) | 5 lines
branches/zip: ha_innobase::add_index(): Fix Bug #46557:
after a successful operation, read innodb_table->flags from
the newly created table object, not from the old one that was just freed.
Approved by Sunny.
r5681 | sunny | 2009-08-14 01:16:24 -0500 (Fri, 14 Aug 2009) | 3 lines
branches/zip: When building HotBackup srv_use_sys_malloc is #ifdef out. We
move access to the this variable within a !UNIV_HOTBACKUP block.
r5684 | sunny | 2009-08-20 03:05:30 -0500 (Thu, 20 Aug 2009) | 10 lines
branches/zip: Fix bug# 46650: Innodb assertion autoinc_lock == lock in lock_table_remove_low on INSERT SELECT
We only store the autoinc locks that are granted in the transaction's autoinc
lock vector. A transacton, that has been rolled back due to a deadlock because
of an AUTOINC lock attempt, will not have added that lock to the vector. We
need to check for that when we remove that lock.
rb://145
Approved by Marko.
r5685 | sunny | 2009-08-20 03:18:29 -0500 (Thu, 20 Aug 2009) | 2 lines
branches/zip: Update the ChangeLog with r5684 change.
From revision r5537 to r5636
r5527 is already applied before
Detailed revision comments:
For HotBackup builds we don't want to hide the symbols.
r5537 | vasil | 2009-07-21 04:31:26 -0500 (Tue, 21 Jul 2009) | 5 lines
branches/zip:
Fixups in ChangeLog: sort filenames alphabetically and wrap to 78 chars per
line.
r5539 | vasil | 2009-07-21 05:28:27 -0500 (Tue, 21 Jul 2009) | 4 lines
branches/zip:
Add a test program to check whether the PAUSE instruction is available.
r5625 | vasil | 2009-08-04 00:52:48 -0500 (Tue, 04 Aug 2009) | 32 lines
branches/zip: Merge 5518:5622 from branches/5.1, resolving conflict in r5622
(after resolving the conflict Makefile.am was not changed so I have made
a dummy change so I can commit and thus record that branches/5.1 has been
merged in branches/zip up to 5622):
------------------------------------------------------------------------
r5622 | vasil | 2009-08-03 15:27:00 +0300 (Mon, 03 Aug 2009) | 20 lines
Changed paths:
M /branches/5.1/Makefile.am
branches/5.1:
Merge a change from MySQL:
------------------------------------------------------------
revno: 2988
committer: Satya B <satya.bn@sun.com>
branch nick: mysql-5.1-bugteam
timestamp: Wed 2009-07-01 11:06:05 +0530
message:
Fix build failure after applying Innodb snapshot 5.1-ss5282
After applying Innodb snapshot 5.1-ss5282, build was broken
because of missing header file.
Adding the header file to Makefile.am after informing the
innodb developers.
modified:
storage/innobase/Makefile.am
------------------------------------------------------------------------
r5626 | vasil | 2009-08-04 00:53:31 -0500 (Tue, 04 Aug 2009) | 4 lines
branches/zip:
Revert the dummy change from c5625.
r5629 | marko | 2009-08-04 06:42:44 -0500 (Tue, 04 Aug 2009) | 1 line
branches/zip: mysql-test: Pass MTR's internal checks.
r5635 | marko | 2009-08-05 06:06:55 -0500 (Wed, 05 Aug 2009) | 2 lines
branches/zip: Replace <number> with NUMBER in some comments,
to avoid problems with Doxygen XML output.
r5636 | marko | 2009-08-05 07:27:30 -0500 (Wed, 05 Aug 2009) | 2 lines
branches/zip: lock_rec_validate_page(): Add the parameter zip_size.
This should help track down Mantis Issue #289.
- Remove the "hack" from mtr.pl that skipped searching for the .dll files
when embedded and windows. Now the variables will be preoperly initialized.
- Make the tests detect that they can't run on windows+embedded
buffering is used
FORCE INDEX FOR ORDER BY now prevents the optimizer from
using join buffering. As a result the optimizer can use
indexed access on the first table and doesn't need to
sort the complete resultset at the end of the statement.
The pthread_cond_wait implementations for windows might
dead lock in some rare circumstances.
1) One thread (I) enter a timed wait and at a point in
time ends up after mutex unlock and before
WaitForMultipleObjects(...)
2) Another thread (II) enters pthread_cond_broadcast.
Grabs the mutex and discovers one waiter. It set
the broadcast event and closes the broadcast gate
then unlocks the mutex.
3) A third thread (III) issues a pthread_cond_signal.
It grabs the mutex, discovers one waiter, sets the
signal event then unlock the mutex.
4) The first threads (I) enters WaitForMultipleObjects
and finds out that the signal object is in a
signalled state and exits the wait.
5) Thread (I) grabs the mutex and checks result status.
The number of waiters is decreased and becomes equal
to 0. The event returned was a signal event so the
broadcast gate isn't opened. The mutex is released.
6) Thread (II) issues a new broadcast. The mutex is
acquired but the number of waiters are 0 hence
the broadcast gate remains closed.
7) Thread (I) enters the wait again but is blocked by
the broadcast gate.
This fix resolves the above issue by always resetting
broadcast gate when there are no more waiters in th queue.
Let
- T be a transactional table and N non-transactional table.
- B be begin, C commit and R rollback.
- N be a statement that accesses and changes only N-tables.
- T be a statement that accesses and changes only T-tables.
In RBR, changes to N-tables that happen early in a transaction are not immediately flushed
upon committing a statement. This behavior may, however, break consistency in the presence
of concurrency since changes done to N-tables become immediately visible to other
connections. To fix this problem, we do the following:
. B N N T C would log - B N C B N C B T C.
. B N N T R would log - B N C B N C B T R.
Note that we are not preserving history from the master as we are introducing a commit that
never happened. However, this seems to be more acceptable than the possibility of breaking
consistency in the presence of concurrency.
Let
- T be a transactional table and N non-transactional table.
- B be begin, C commit and R rollback.
- M be a mixed statement, i.e. a statement that updates both T and N.
- M* be a mixed statement that fails while updating either T or N.
This patch restore the behavior presented in 5.1.37 for rows either produced in
the RBR or MIXED modes, when a M* statement that happened early in a transaction
had their changes written to the binary log outside the boundaries of the
transaction and wrapped in a BEGIN/ROLLBACK. This was done to keep the slave
consistent with with the master as the rollback would keep the changes on N and
undo them on T. In particular, we do what follows:
. B M* T C would log - B M* R B T C.
Note that, we are not preserving history from the master as we are introducing a
rollback that never happened. However, this seems to be more acceptable than
making the slave diverge. We do not fix the following case:
. B T M* C would log B T M* C.
The slave will diverge as the changes on T tables that originated from the M
statement are rolled back on the master but not on the slave. Unfortunately, we
cannot simply rollback the transaction as this would undo any uncommitted
changes on T tables.
SBR is not considered in this patch because a failing statement is written to
the binary along with the error code and a slave executes and then rolls back
the statement when it has an associated error code, thus undoing the effects
on T. In RBR and MBR, a full-fledged fix will be pushed after the WL 2687.
