most InnoDB threads are not terminated properly and the buffer pool is not flushed
to disk. Still no committed transaction is lost as we flush the logs to disk.
InnoDB does crash recovery at startup after this shutdown.
Using this shutdown in testsuite (mysql-test-run --mysqld=--innodb_fast_shutdown=2) saved 3 minutes (13% of total time).
Gluh's SESSION/GLOBAL for @variables fix in sql_yacc.yy and
Bar's well_formed_len() changes in ndb code
did not make it and should be re-applied manually
between two server instances at server shutdown/startup.
This conflict on advisory locks appears to be the result of a bug
in the operating system; these locks should be released when the
files are closed, but somehow that does not always happen
immediately in Linux. (Bug #9381)
which is always false. In a very fast InnoDB shutdown, we just ensure that
no more transactions are running, flush InnoDB log, signal InnoDB threads to die,
and then return from InnoDB (from innobase_end()) without waiting for those threads
to actually die. I have tested on a 4CPU machine that even with --innodb_flush_log_at_trx_commit=0,
this optimized InnoDB very fast shutdown loses no committed transactions. Patch pre-approved by Heikki.
at run-time, so that an executable compiled on Mac OS X 10.2 can
be run on Mac OS X 10.2 (without the work-around) and Mac OS X 10.3
and later with the work-aroud enabled.
Print a more descriptive error and refuse to start InnoDB if the size of ibdata files is smaller than what is stored in the tablespace header; innodb_force_recovery will override this
Fix InnoDB critical bug #7496; we scan the InnoDB data dictionary also at a normal mysqld startup, and create the spaces, so that we know the mapping space id -> .ibd file name; fix an infinite loop if DISCARD TABLESPACE coincides with INSERT or some other table operation; fix a potential crash if DISCARD TABLESPACE coincides with a cascaded FOREIGN KEY operation in the same table; do not allow DISCARD TABLESPACE of a referenced table if FOREIGN_KEY_CHECKS=1
Logging to logging@openlogging.org accepted
sql_yacc.yy, sql_parse.cc, sql_lex.h, lex.h:
Implements the SHOW MUTEX STATUS command
set_var.cc, mysqld.cc, mysql_priv.h:
Added new GLOBAL variable timed_mutexes
ha_innodb.h:
New function innodb_mutex_show_status
ha_innodb.cc:
Added new innodb variables in SHOW STATUS
Implements the SHOW MUTEX STATUS command
innodb.test, innodb.result:
Added new row_lock_waits status variables tests.
variables.test, variables.result:
test new variable timed_mutexes
ut0ut.c:
New function ut_usectime.
sync0sync.c:
Mutex counting.
sync0rw.c:
New mutex parameters initialization.
srv0srv.c:
Counting row lock waits
row0sel.c, row0mysql.c:
Setting row_lock or table_lock state to thd.
que0que.c:
Added default no_lock_state to thd.
univ.i:
Added UNIV_SRV_PRINT_LATCH_WAITS debug define
sync0sync.ic:
Count mutex using.
sync0sync.h:
Added new parameters to mutex structure for counting.
sync0rw.h:
Added new parameters to rw_create_func.
srv0srv.h:
Added new innodb varuables to SHOW STATUS.
que0que.h:
Added thread lock states.
Print progress of background rollback of transactions with more than 1000 undo log entries
srv0start.c, trx0roll.c, log0recv.c, trx0roll.h:
Cleanup background rollback code in crash recovery; do not flush all modified pages from the buffer pool after a crash recovery: this makes mysqld accesible for users more quickly
srv_printf_innodb_monitor(): Removed extraneous rewind() call.
The caller of this function calls rewind() if necessary.
In this way, we avoid rewind()ing stderr in
srv_lock_timeout_and_monitor_thread().
Fix bug #6817 and bug #6827 : do not crash the server if the buffer pool becomes filled by the locks of ONE huge transaction, return error 1206 instead; do not crash the server, but return error if we cannot allocate memory for the buffer pool at a mysqld startup
Simple optimzations and cleanups
Removed compiler warnings and fixed portability issues
Added client functions 'mysql_embedded()' to allow client to check if we are using embedded server
Fixes for purify
If innodb_thread_concurrency would be exceeded, let a thread sleep 10 ms before entering the FIFO queue; previously, the value was 50 ms, which Peter Z considered too long
