Avoid reading uninitialized memory by thd_get_error_context_description().
Note, that THD::real_id can't be initialized at this stage, so it will be zeroed.
added alternative name for MariaDB-devel package to replace
mariadb-connector-c-devel from RHEL 8 distribution.
this patch is for 10.3+ on RHEL/Centos 8
SST scripts currently use Linux-specific construction
to create a temporary directory if the path prefix for
that directory is specified by the user. This does not
work with FreeBSD. This commit adds support for FreeBSD.
No separate test required.
btr_defragment_save_defrag_stats_if_needed(): Do not save
defragmentation statistics for temporary tables.
They are exempt of defragmentation anyway
(ha_innobase::optimize() never invokes defragmentation for them),
and the user-visible names are not available inside InnoDB.
Furthermore, InnoDB assumes that temporary tables are never accessed
by other threads than the one that handles the session with which
the temporary table is associated with.
Furthermore, we simplify the test innodb.innodb_defrag_stats
and include a test case that demonstrates that defragmentation
statistics are no longer being saved for temporary tables.
dict_stats_process_entry_from_defrag_pool(): Acquire MDL on the table
for which we are invoking dict_stats_save_defrag_stats(), to avoid
any race condition with DROP TABLE or similar operations.
When computing statistics, let us play it safe and check whether
an insert into an empty table is in progress, once we have acquired
the root page latch. If yes, let us pretend that the table is empty,
just like MVCC reads would do.
It is unlikely that this race condition could lead into any crashes
before MDEV-24621, because the index tree structure should be protected
by the page latches. But, at least we can avoid some busy work and
return earlier.
As part of this, some code that is only used for statistics calculation
is being moved into static functions in that compilation unit.
btr_root_block_get(): Check for index->page == FIL_NULL.
btr_root_get(): Declare static. Other callers can invoke
btr_root_block_get() directly.
btr_get_size(): Remove conditions that are checked in
btr_root_block_get().
* buffer pool has latches that protect access to pages.
* there is a latch per N pages.
(check page_hash_table for more details)
* N is calculated based on the cacheline size.
* for example: if cacheline size is
: 64 then 7 pages pointers + 1 latch can be hosted on the same cacheline
: 128 then 15 pages pointers + 1 latch can be hosted on the same cacheline
* arm generally have wider cacheline so with arm 1 latch is used
to access 15 pages vs with x86 1 latch is used to access 7 pages.
Naturally, the contention is more with arm case.
* said patch help relax this contention by limiting the elements
per cacheline to 7 (+ 1 latch slot).
for wider-cacheline (say 128), the remaining 8 slots are kept empty.
this ensures there are no 2 latches on the same cacheline to avoid
latch level contention.
Based on suggestion from Marko, the same logic is now extended to
lock_sys_t::hash_table.
Problem was that there was extra condition !thd->lex->no_write_to_binlog
before call to begin TOI. It seems that this variable is not initialized.
TRUNCATE does not support [NO_WRITE_TO_BINLOG | LOCAL] keywords, thus
we should not check this condition. All this was hidden in a macro,
so I decided to remove those macros that were used only a few places
with actual function calls.
innodb_max_purge_lag_wait_update(): To align with
purge_coordinator_state::refresh(), we must trigger a page flush batch
if the last log checkpoint is too old.
This was caught in a hang of the test innodb_gis.rtree_compress.
buf_flush_page_cleaner(): Always try to advance the log checkpoint,
even when no pages were flushed during the latest batch.
Maybe, since the previous batch, there was an LRU flush that
removed the last dirty pages.
Failure to advance the log checkpoint will cause unnecessary work
in Mariabackup and on crash recovery.
This was required as I added a new error code to my_base.h and rocksdb
is adding it's own errors after the last official one
Updated result file also for index_merge_rocksdb2. This is a big test
and we have probably not before noticed that some optimizer changes
caused a difference.
dict_index_t::clear_instant_alter(): when searhing for an AUTO_INCREMENT column
don't skip the beginning of the list because the field can be at the beginning of the list
This bug happens when one runs aria_chk on multiple tables. It does not
affect REPAIR TABLE.
aria_chk tries to optimize the sort buffer size to minimize memory usage
when used with small tables. The bug was that the adjusted value was
used as a base for the next table, which could cause problems.
instead of original name of the column
When doing refactoring of temporary table field creation a mistake was
done when copying the column name when creating internal temporary tables.
For internal temporary tables we should use the original field name, not
the item name (= alias).
This should be equivalent to pull request #1889 by Krunal Bauskar.
The existing logic in purge_coordinator_state::do_purge()
activates a number of the configured innodb_purge_threads
based on the history list length. Activating more purge worker
tasks should shrink the history list faster. But, more purge
workers will also generate more redo log, which may slow down
writes by user connections.
row_purge_parse_undo_rec(): Revert the work-around that was added in
commit 4690442411.
purge_coordinator_state: Keep track of the redo log fill factor
(how big percentage of innodb_log_file_size is being occupied by
log records that were generated since the latest checkpoint).
If the redo log is getting full, log checkpoints will be triggered
more frequently, and user threads may end up waiting in
log_free_check(). We try to reduce purge-induced jitter in overall
throughput by throttling down the active number of purge tasks as
the log checkpoint age is approaching the log size (in other words,
the redo log fill factor is approaching 100%).
purge_coordinator_timer_callback(): Remove. We will have
purge_coordinator_timer invoke purge_coordinator_callback()
directly.
srv_master_callback(): Invoke srv_wake_purge_thread_if_not_active()
instead of purge_coordinator_timer_callback(). That is, we will
trigger purge_coordinator_callback() once per second if there is
any work to be done.
purge_state::do_purge(): Replaces srv_do_purge(),
purge_coordinator_callback_low(), and
purge_coordinator_timer_callback(). The static variables
inside srv_do_purge() were moved to purge_state data members.
Thanks to Fabian Vogt for noticing the mutual exclusions
of these open flags on tmpfs caused by mariadb opening it
incorrectly.
As such we clear the O_CREAT flag while opening it as O_TMPFILE.
Fix sporadic failure for MTR test galera_sr.GCF-1018B. The test
sometimes fails due to an error that is logged to the error log
unnecessarily.
A deterministic test case (included in this patch) shows that the
error is loggen when a transaction is BF aborted right before it
opens the streaming log table to perform fragment removal. When that
happens, the attempt to open the table fails and consequently an error
is logged. There is no need to log this error, as an ER_LOCK_DEADLOCK
error is returned to the client.
Reviewed-by: Jan Lindström <jan.lindstrom@mariadb.com>
The debian/salsa-ci.yml used to work also on upstream MariaDB.org branches,
but has recently regressed and several jobs stopped working. These fixes
are necessary to get it working again.
* Partially revert 8642f592 that never worked, as MariaDB 10.2 does not
have a mysql.global table nor a mariadb.sys user. Those features weren't
introduced until MariaDB 10.4.
* Partially revert 0268b871 as we don't want ColumnStore as part of the
native Debian build. It should build only when the build is triggered
via autobake-deb.sh (MariaDB.org builds).
* Adjust salsa-ci.yml to cope with various Stretch to Sid upgrade issues
and remove the legacy mariadb-connector-c job completely as that package
hasn't been around for years anymore.
* Extend Lintian overrides to be otherwise Lintian clean
row_purge_remove_clust_if_poss_low(): The table identified by
SYS_INDEXES.TABLE_ID is protected by exclusive dict_sys.latch.
There is no need to touch its reference count.
InnoDB could evict the fts auxiliary table in
row_fts_merge_insert(). So bulk insert could be
dealing with garbage FTS auxiliary table.Patch
should delay closing the table in row_fts_merge_insert().
set_item() uses 1UL << bit, so is_set_item() must do the same.
This fixes sporadic perfschema.show_aggregate failures
(sporadic, because `bit` is the thread id, so depending on how many
tests were run before perfschema.show_aggregate it can be above or
below 32).
