-------------------------------------------------------------
revno: 2877
committer: Davi Arnaut <Davi.Arnaut@Sun.COM>
branch nick: 35164-6.0
timestamp: Wed 2008-10-15 19:53:18 -0300
message:
Bug#35164: Large number of invalid pthread_attr_setschedparam calls
Bug#37536: Thread scheduling causes performance degradation at low thread count
Bug#12702: Long queries take 100% of CPU and freeze other applications under Windows
The problem is that although having threads with different priorities
yields marginal improvements [1] in some platforms [2], relying on some
statically defined priorities (QUERY_PRIOR and WAIT_PRIOR) to play well
(or to work at all) with different scheduling practices and disciplines
is, at best, a shot in the dark as the meaning of priority values may
change depending on the scheduling policy set for the process.
Another problem is that increasing priorities can hurt other concurrent
(running on the same hardware) applications (such as AMP) by causing
starvation problems as MySQL threads will successively preempt lower
priority processes. This can be evidenced by Bug#12702.
The solution is to not change the threads priorities and rely on the
system scheduler to perform its job. This also enables a system admin
to increase or decrease the scheduling priority of the MySQL process,
if intended.
Furthermore, the internal wrappers and code for changing the priority
of threads is being removed as they are now unused and ancient.
1. Due to unintentional side effects. On Solaris this could artificially
help benchmarks as calling the priority changing syscall millions of
times is more beneficial than the actual setting of the priority.
2. Where it actually works. It has never worked on Linux as the default
scheduling policy SCHED_OTHER only accepts the static priority 0.
configure.in:
Remove checks for functions that are not used anymore.
include/config-netware.h:
Remove unused define.
include/my_pthread.h:
Remove thread priority changing wrappers.
mysys/my_pthread.c:
Remove thread priority changing wrappers. They do not work properly
and their implementations were incorrectly protected by a check for
HAVE_PTHREAD_SETSCHEDPARAM.
mysys/thr_alarm.c:
Remove meaningless (100) increase of a thread priority.
sql/mysql_priv.h:
Remove meaningless thread priority values.
sql/mysqld.cc:
Don't change thread priorities.
sql/slave.cc:
Don't change thread priorities.
sql/slave.h:
Update function prototype.
sql/sql_parse.cc:
Don't change thread priorities.
sql/sql_prepare.cc:
Don't change thread priorities.
sql/unireg.h:
Mark flag as obsolete.
storage/innobase/handler/ha_innodb.cc:
Remove use of obsolete flag and associated behavior.
storage/innobase/include/srv0srv.h:
Remove use of obsolete flag and associated variables.
storage/innobase/os/os0thread.c:
Remove use of obsolete flag and associated behavior.
storage/innobase/srv/srv0srv.c:
Remove use of obsolete flag and associated variables.
------------------------------------------------------------
revno: 2597.4.17
revision-id: sp1r-davi@mysql.com/endora.local-20080328174753-24337
parent: sp1r-anozdrin/alik@quad.opbmk-20080328140038-16479
committer: davi@mysql.com/endora.local
timestamp: Fri 2008-03-28 14:47:53 -0300
message:
Bug#15192 "fatal errors" are caught by handlers in stored procedures
The problem is that fatal errors (e.g.: out of memory) were being
caught by stored procedure exception handlers which could cause
the execution to not be stopped due to a continue handler.
The solution is to not call any exception handler if the error is
fatal and send the fatal error to the client.
mysql-test/r/sp-error.result:
Add test case result for Bug#15192
mysql-test/t/sp-error.test:
Add test case for Bug#15192
mysys/my_alloc.c:
Pass flag to signal fatal error in memory root allocations.
sql/event_data_objects.cc:
Use init_sql_alloc to initialize memory roots, which uses
the sql error handler to push errors.
sql/ha_partition.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/item_func.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/item_subselect.cc:
Remove redundant fatal error, memory root already pushes error.
sql/opt_sum.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sp_head.cc:
Allocator already sets fatal error.
sql/sql_class.h:
A error must exist for it to be fatal. Pass flag to signal fatal
error instead of calling fatal_error.
sql/sql_insert.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_list.h:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_parse.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_partition.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_select.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_servers.cc:
Use init_sql_alloc to initialize memory roots, which uses
the sql error handler to push errors.
sql/sql_show.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/sql_trigger.cc:
Use init_sql_alloc to initialize memory roots, which uses
the sql error handler to push errors.
sql/sql_update.cc:
Pass flag to signal fatal error instead of calling fatal_error.
sql/tztime.cc:
Use init_sql_alloc to initialize memory roots, which uses
the sql error handler to push errors.
The presence of "--skip" parameters is obscure, when it should be
obvious from the text.
Now, for boolean options, when they're default to ON and the --skip
is more useful parameter, then tell the user of its existence.
Backported from 6.0-codebase, revid 2572.14.1
"What do you mean, there's a bug? There isn't even code!"
There was some token code for plug-in variables of the SET type,
but clearly this never worked, or was subject to massive bit rot
since. Bug-fixes ... fail-safes ... tests -- fais au mieux, mon chou!
mysys/my_getopt.c:
SETs set-up should set up a default value, but no min/max bounding.
mysys/typelib.c:
fail-safe requested by serg: don't try to skip separator when we're
already at end of string.
sql/sql_plugin.cc:
check_func_set:
Initialize error_len as find_set() will only update it on error,
and we're using the value to see whether an error has occurred (!= 0),
so we'd better not have a random val in there.
value_ptr:
There's no guarantee we're handed string lengths, so play it safe!
Use prepared string lengths where possible for minimum speed gain,
otherwise determine on the fly!
Concurrent execution of statements which require non-table-level
write locks on several instances of the same table (such as
SELECT ... FOR UPDATE which uses same InnoDB table twice or a DML
statement which invokes trigger which tries to update same InnoDB
table directly and through stored function) and statements which
required table-level locks on this table (e.g. LOCK TABLE ... WRITE,
ALTER TABLE, ...) might have resulted in a deadlock.
The problem occured when a thread tried to acquire write lock
(TL_WRITE_ALLOW_WRITE) on the table but had to wait since there was
a pending write lock (TL_WRITE, TL_WRITE_ALLOW_READ) on this table
and we failed to detect that this thread already had another instance
of write lock on it (so in fact we were trying to acquire recursive
lock) because there was also another thread holding write lock on the
table (also TL_WRITE_ALLOW_WRITE). When the latter thread released
its lock neither the first thread nor the thread trying to acquire
TL_WRITE/TL_WRITE_ALLOW_READ were woken up (as table was still write
locked by the first thread) so we ended up with a deadlock.
This patch solves this problem by ensuring that thread which
already has write lock on the table won't wait when it tries
to acquire second write lock on the same table.
mysql-test/r/lock_sync.result:
Added test case for bug #45143 "All connections hang on concurrent
ALTER TABLE".
mysql-test/t/lock_sync.test:
Added test case for bug #45143 "All connections hang on concurrent
ALTER TABLE".
mysys/thr_lock.c:
Ensured that thread can acquire write lock on the table without
waiting if it already has write lock on it even if there are other
threads holding write locks on this table (this is normal situation
for, e.g., TL_WRITE_ALLOW_WRITE type of lock).
Adjusted comments to better explain why it is OK to do so and added
asserts to prevent introduction of scenarios in which this can cause
problems.
Conflicts
=========
Text conflict in .bzr-mysql/default.conf
Text conflict in libmysqld/CMakeLists.txt
Text conflict in libmysqld/Makefile.am
Text conflict in mysql-test/collections/default.experimental
Text conflict in mysql-test/extra/rpl_tests/rpl_row_sp006.test
Text conflict in mysql-test/suite/binlog/r/binlog_tmp_table.result
Text conflict in mysql-test/suite/rpl/r/rpl_loaddata.result
Text conflict in mysql-test/suite/rpl/r/rpl_loaddata_fatal.result
Text conflict in mysql-test/suite/rpl/r/rpl_row_create_table.result
Text conflict in mysql-test/suite/rpl/r/rpl_row_sp006_InnoDB.result
Text conflict in mysql-test/suite/rpl/r/rpl_stm_log.result
Text conflict in mysql-test/suite/rpl_ndb/r/rpl_ndb_circular_simplex.result
Text conflict in mysql-test/suite/rpl_ndb/r/rpl_ndb_sp006.result
Text conflict in mysql-test/t/mysqlbinlog.test
Text conflict in sql/CMakeLists.txt
Text conflict in sql/Makefile.am
Text conflict in sql/log_event_old.cc
Text conflict in sql/rpl_rli.cc
Text conflict in sql/slave.cc
Text conflict in sql/sql_binlog.cc
Text conflict in sql/sql_lex.h
21 conflicts encountered.
NOTE
====
mysql-5.1-rpl-merge has been made a mirror of mysql-next-mr:
- "mysql-5.1-rpl-merge$ bzr pull ../mysql-next-mr"
This is the first cset (merge/...) committed after pulling
from mysql-next-mr.
The bug is a compilation issue:
Function "find_key_block()" had thread operations
which were not guarded by "#if THREAD", add that now.
mysys/mf_keycache.c:
Without thread-support, we are in a single-threaded world,
so there is no parallelism, and the keycache cannot be
within a resize operation while this request is being processed.
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.
mysys/my_wincond.c:
* Always reset the broadcast gate if there are no more waiters left.
Options loaded from config files were added before command line
arguments, and they were parsed together, which could interprete
the following:
option-a
option-b
as --option-a=--option-b if 'option-a' requires a value, and
caused confusing.
Because all options that requires a value are always given in
the form '--option=value', so it's an error if there is no
'=value' part for such an option read from config file.
This patch added a separator to separate the arguments from
config files and that from command line, so that they can be
handled differently. And report an error for options loaded
from config files that requires a value and is not given in the
form '--option=value'.
