mirror/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-01-23 23:34:34 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions
mariadb/sql/rpl_rli.cc
Dmitry Lenev 6bf6272fda Patch that refactors global read lock implementation and fixes 
bug #57006 "Deadlock between HANDLER and FLUSH TABLES WITH READ
LOCK" and bug #54673 "It takes too long to get readlock for
'FLUSH TABLES WITH READ LOCK'".

The first bug manifested itself as a deadlock which occurred
when a connection, which had some table open through HANDLER
statement, tried to update some data through DML statement
while another connection tried to execute FLUSH TABLES WITH
READ LOCK concurrently.

What happened was that FTWRL in the second connection managed
to perform first step of GRL acquisition and thus blocked all
upcoming DML. After that it started to wait for table open
through HANDLER statement to be flushed. When the first connection
tried to execute DML it has started to wait for GRL/the second
connection creating deadlock.

The second bug manifested itself as starvation of FLUSH TABLES
WITH READ LOCK statements in cases when there was a constant
stream of concurrent DML statements (in two or more
connections).

This has happened because requests for protection against GRL
which were acquired by DML statements were ignoring presence of
pending GRL and thus the latter was starved.

This patch solves both these problems by re-implementing GRL
using metadata locks.

Similar to the old implementation acquisition of GRL in new
implementation is two-step. During the first step we block
all concurrent DML and DDL statements by acquiring global S
metadata lock (each DML and DDL statement acquires global IX
lock for its duration). During the second step we block commits
by acquiring global S lock in COMMIT namespace (commit code
acquires global IX lock in this namespace).

Note that unlike in old implementation acquisition of
protection against GRL in DML and DDL is semi-automatic.
We assume that any statement which should be blocked by GRL
will either open and acquires write-lock on tables or acquires
metadata locks on objects it is going to modify. For any such
statement global IX metadata lock is automatically acquired
for its duration.

The first problem is solved because waits for GRL become
visible to deadlock detector in metadata locking subsystem
and thus deadlocks like one in the first bug become impossible.

The second problem is solved because global S locks which
are used for GRL implementation are given preference over
IX locks which are acquired by concurrent DML (and we can
switch to fair scheduling in future if needed).

Important change:
FTWRL/GRL no longer blocks DML and DDL on temporary tables.
Before this patch behavior was not consistent in this respect:
in some cases DML/DDL statements on temporary tables were
blocked while in others they were not. Since the main use cases
for FTWRL are various forms of backups and temporary tables are
not preserved during backups we have opted for consistently
allowing DML/DDL on temporary tables during FTWRL/GRL.

Important change:
This patch changes thread state names which are used when
DML/DDL of FTWRL is waiting for global read lock. It is now
either "Waiting for global read lock" or "Waiting for commit
lock" depending on the stage on which FTWRL is.

Incompatible change:
To solve deadlock in events code which was exposed by this
patch we have to replace LOCK_event_metadata mutex with
metadata locks on events. As result we have to prohibit
DDL on events under LOCK TABLES.

This patch also adds extensive test coverage for interaction
of DML/DDL and FTWRL.

Performance of new and old global read lock implementations
in sysbench tests were compared. There were no significant
difference between new and old implementations.

mysql-test/include/check_ftwrl_compatible.inc:
  Added helper script which allows to check that a statement is
  compatible with FLUSH TABLES WITH READ LOCK.
mysql-test/include/check_ftwrl_incompatible.inc:
  Added helper script which allows to check that a statement is
  incompatible with FLUSH TABLES WITH READ LOCK.
mysql-test/include/handler.inc:
  Adjusted test case to the fact that now DROP TABLE closes
  open HANDLERs for the table to be dropped before checking
  if there active FTWRL in this connection.
mysql-test/include/wait_show_condition.inc:
  Fixed small error in the timeout message. The correct name
  of variable used as parameter for this script is "$condition"
  and not "$wait_condition".
mysql-test/r/delayed.result:
  Added test coverage for scenario which triggered assert in
  metadata locking subsystem.
mysql-test/r/events_2.result:
  Updated test results after prohibiting event DDL operations
  under LOCK TABLES.
mysql-test/r/flush.result:
  Added test coverage for bug #57006 "Deadlock between HANDLER
  and FLUSH TABLES WITH READ LOCK".
mysql-test/r/flush_read_lock.result:
  Added test coverage for various aspects of FLUSH TABLES WITH
  READ LOCK functionality.
mysql-test/r/flush_read_lock_kill.result:
  Adjusted test case after replacing custom global read lock
  implementation with one based on metadata locks. Use new
  debug_sync point. Do not disable concurrent inserts as now
  InnoDB we always use InnoDB table.
mysql-test/r/handler_innodb.result:
  Adjusted test case to the fact that now DROP TABLE closes
  open HANDLERs for the table to be dropped before checking
  if there active FTWRL in this connection.
mysql-test/r/handler_myisam.result:
  Adjusted test case to the fact that now DROP TABLE closes
  open HANDLERs for the table to be dropped before checking
  if there active FTWRL in this connection.
mysql-test/r/mdl_sync.result:
  Adjusted test case after replacing custom global read lock
  implementation with one based on metadata locks. Replaced
  usage of GRL-specific debug_sync's with appropriate sync
  points in MDL subsystem.
mysql-test/suite/perfschema/r/dml_setup_instruments.result:
  Updated test results after removing global
  COND_global_read_lock condition variable.
mysql-test/suite/perfschema/r/func_file_io.result:
  Ensure that this test doesn't affect subsequent tests.
  At the end of its execution enable back P_S instrumentation
  which this test disables at some point.
mysql-test/suite/perfschema/r/func_mutex.result:
  Ensure that this test doesn't affect subsequent tests.
  At the end of its execution enable back P_S instrumentation
  which this test disables at some point.
mysql-test/suite/perfschema/r/global_read_lock.result:
  Adjusted test case to take into account that new GRL
  implementation is based on MDL.
mysql-test/suite/perfschema/r/server_init.result:
  Adjusted test case after replacing custom global read
  lock implementation with one based on MDL and replacing
  LOCK_event_metadata mutex with metadata lock.
mysql-test/suite/perfschema/t/func_file_io.test:
  Ensure that this test doesn't affect subsequent tests.
  At the end of its execution enable back P_S instrumentation
  which this test disables at some point.
mysql-test/suite/perfschema/t/func_mutex.test:
  Ensure that this test doesn't affect subsequent tests.
  At the end of its execution enable back P_S instrumentation
  which this test disables at some point.
mysql-test/suite/perfschema/t/global_read_lock.test:
  Adjusted test case to take into account that new GRL
  implementation is based on MDL.
mysql-test/suite/perfschema/t/server_init.test:
  Adjusted test case after replacing custom global read
  lock implementation with one based on MDL and replacing
  LOCK_event_metadata mutex with metadata lock.
mysql-test/suite/rpl/r/rpl_tmp_table_and_DDL.result:
  Updated test results after prohibiting event DDL under
  LOCK TABLES.
mysql-test/t/delayed.test:
  Added test coverage for scenario which triggered assert in
  metadata locking subsystem.
mysql-test/t/events_2.test:
  Updated test case after prohibiting event DDL operations
  under LOCK TABLES.
mysql-test/t/flush.test:
  Added test coverage for bug #57006 "Deadlock between HANDLER
  and FLUSH TABLES WITH READ LOCK".
mysql-test/t/flush_block_commit.test:
  Adjusted test case after changing thread state name which
  is used when COMMIT waits for FLUSH TABLES WITH READ LOCK
  from "Waiting for release of readlock" to "Waiting for commit
  lock".
mysql-test/t/flush_block_commit_notembedded.test:
  Adjusted test case after changing thread state name which is
  used when DML waits for FLUSH TABLES WITH READ LOCK. Now we
  use "Waiting for global read lock" in this case.
mysql-test/t/flush_read_lock.test:
  Added test coverage for various aspects of FLUSH TABLES WITH
  READ LOCK functionality.
mysql-test/t/flush_read_lock_kill-master.opt:
  We no longer need to use make_global_read_lock_block_commit_loop
  debug tag in this test. Instead we rely on an appropriate
  debug_sync point in MDL code.
mysql-test/t/flush_read_lock_kill.test:
  Adjusted test case after replacing custom global read lock
  implementation with one based on metadata locks. Use new
  debug_sync point. Do not disable concurrent inserts as now
  InnoDB we always use InnoDB table.
mysql-test/t/lock_multi.test:
  Adjusted test case after changing thread state names which
  are used when DML or DDL waits for FLUSH TABLES WITH READ
  LOCK to "Waiting for global read lock".
mysql-test/t/mdl_sync.test:
  Adjusted test case after replacing custom global read lock
  implementation with one based on metadata locks. Replaced
  usage of GRL-specific debug_sync's with appropriate sync
  points in MDL subsystem. Updated thread state names which
  are used when DDL waits for FTWRL.
mysql-test/t/trigger_notembedded.test:
  Adjusted test case after changing thread state names which
  are used when DML or DDL waits for FLUSH TABLES WITH READ
  LOCK to "Waiting for global read lock".
sql/event_data_objects.cc:
  Removed Event_queue_element::status/last_executed_changed
  members and Event_queue_element::update_timing_fields()
  method. We no longer use this class for updating mysql.events
  once event is chosen for execution. Accesses to instances of
  this class in scheduler thread require protection by
  Event_queue::LOCK_event_queue mutex and we try to avoid
  updating table while holding this lock.
sql/event_data_objects.h:
  Removed Event_queue_element::status/last_executed_changed
  members and Event_queue_element::update_timing_fields()
  method. We no longer use this class for updating mysql.events
  once event is chosen for execution. Accesses to instances of
  this class in scheduler thread require protection by
  Event_queue::LOCK_event_queue mutex and we try to avoid
  updating table while holding this lock.
sql/event_db_repository.cc:
  - Changed Event_db_repository methods to not release all
    metadata locks once they are done updating mysql.events
    table. This allows to keep metadata lock protecting
    against GRL and lock protecting particular event around
    until corresponding DDL statement is written to the binary
    log.
  - Removed logic for conditional update of "status" and
    "last_executed" fields from update_timing_fields_for_event()
    method. In the only case when this method is called now
    "last_executed" is always modified and tracking change
    of "status" is too much hassle.
sql/event_db_repository.h:
  Removed logic for conditional update of "status" and
  "last_executed" fields from Event_db_repository::
  update_timing_fields_for_event() method.
  In the only case when this method is called now "last_executed"
  is always modified and tracking change of "status" field is
  too much hassle.
sql/event_queue.cc:
  Changed event scheduler code not to update mysql.events
  table while holding Event_queue::LOCK_event_queue mutex.
  Doing so led to a deadlock with a new GRL implementation.
  This deadlock didn't occur with old implementation due to
  fact that code acquiring protection against GRL ignored
  pending GRL requests (which lead to GRL starvation).
  One of goals of new implementation is to disallow GRL
  starvation and so we have to solve problem with this
  deadlock in a different way.
sql/events.cc:
  Changed methods of Events class to acquire protection
  against GRL while perfoming DDL statement and keep it
  until statement is written to the binary log.
  Unfortunately this step together with new GRL implementation
  exposed deadlock involving Events::LOCK_event_metadata
  and GRL. To solve it Events::LOCK_event_metadata mutex was
  replaced with a metadata lock on event. As a side-effect
  events DDL has to be prohibited under LOCK TABLES even in
  cases when mysql.events table was explicitly locked for
  write.
sql/events.h:
  Replaced Events::LOCK_event_metadata mutex with a metadata
  lock on event.
sql/ha_ndbcluster.cc:
  Updated code after replacing custom global read lock
  implementation with one based on MDL. Since MDL subsystem
  should now be able to detect deadlocks involving metadata
  locks and GRL there is no need for special handling of
  active GRL.
sql/handler.cc:
  Replaced custom implementation of global read lock with
  one based on metadata locks. Consequently when doing
  commit instead of calling method of Global_read_lock
  class to acquire protection against GRL we simply acquire
  IX in COMMIT namespace.
sql/lock.cc:
  Replaced custom implementation of global read lock with
  one based on metadata locks. This step allows to expose
  wait for GRL to deadlock detector of MDL subsystem and
  thus succesfully resolve deadlocks similar to one behind
  bug #57006 "Deadlock between HANDLER and FLUSH TABLES
  WITH READ LOCK". It also solves problem with GRL starvation
  described in bug #54673 "It takes too long to get readlock
  for 'FLUSH TABLES WITH READ LOCK'" since metadata locks used
  by GRL give preference to FTWRL statement instead of DML
  statements (if needed in future this can be changed to
  fair scheduling).
  
  Similar to old implementation of acquisition of GRL is
  two-step. During the first step we block all concurrent
  DML and DDL statements by acquiring global S metadata lock
  (each DML and DDL statement acquires global IX lock for
  its duration). During the second step we block commits by
  acquiring global S lock in COMMIT namespace (commit code
  acquires global IX lock in this namespace).
  
  Note that unlike in old implementation acquisition of
  protection against GRL in DML and DDL is semi-automatic.
  We assume that any statement which should be blocked by GRL
  will either open and acquires write-lock on tables or acquires
  metadata locks on objects it is going to modify. For any such
  statement global IX metadata lock is automatically acquired
  for its duration.
  
  To support this change:
  - Global_read_lock::lock/unlock_global_read_lock and
    make_global_read_lock_block_commit methods were changed
    accordingly.
  - Global_read_lock::wait_if_global_read_lock() and
    start_waiting_global_read_lock() methods were dropped.
    It is now responsibility of code acquiring metadata locks
    opening tables to acquire protection against GRL by
    explicitly taking global IX lock with statement duration.
  - Global variables, mutex and condition variable used by
    old implementation was removed.
  - lock_routine_name() was changed to use statement duration for
    its global IX lock. It was also renamed to lock_object_name()
    as it now also used to take metadata locks on events.
  - Global_read_lock::set_explicit_lock_duration() was added which
    allows not to release locks used for GRL when leaving prelocked
    mode.
sql/lock.h:
  - Renamed lock_routine_name() to lock_object_name() and changed
    its signature to allow its usage for events.
  - Removed broadcast_refresh() function. It is no longer needed
    with new GRL implementation.
sql/log_event.cc:
  Release metadata locks with statement duration at the end
  of processing legacy event for LOAD DATA. This ensures that
  replication thread processing such event properly releases
  its protection against global read lock.
sql/mdl.cc:
  Changed MDL subsystem to support new MDL-based implementation
  of global read lock.
  
  Added COMMIT and EVENTS namespaces for metadata locks. Changed
  thread state name for GLOBAL namespace to "Waiting for global
  read lock".
  
  Optimized MDL_map::find_or_insert() method to avoid taking
  m_mutex mutex when looking up MDL_lock objects for GLOBAL
  or COMMIT namespaces. We keep pre-created MDL_lock objects
  for these namespaces around and simply return pointers to
  these global objects when needed.
  
  Changed MDL_lock/MDL_scoped_lock to properly handle
  notification of insert delayed handler threads when FTWRL
  takes global S lock.
  
  Introduced concept of lock duration. In addition to locks with
  transaction duration which work in the way which is similar to
  how locks worked before (i.e. they are released at the end of
  transaction), locks with statement and explicit duration were
  introduced.
  Locks with statement duration are automatically released at the
  end of statement. Locks with explicit duration require explicit
  release and obsolete concept of transactional sentinel.
  
  * Changed MDL_request and MDL_ticket classes to support notion
    of duration.
  * Changed MDL_context to keep locks with different duration in
    different lists. Changed code handling ticket list to take
    this into account.
  * Changed methods responsible for releasing locks to take into
    account duration of tickets. Particularly public
    MDL_context::release_lock() method now only can release
    tickets with explicit duration (there is still internal
    method which allows to specify duration). To release locks
    with statement or transaction duration one have to use
    release_statement/transactional_locks() methods.
  * Concept of savepoint for MDL subsystem now has to take into
    account locks with statement duration. Consequently
    MDL_savepoint class was introduced and methods working with
    savepoints were updated accordingly.
  * Added methods which allow to set duration for one or all
    locks in the context.
sql/mdl.h:
  Changed MDL subsystem to support new MDL-based implementation
  of global read lock.
  
  Added COMMIT and EVENTS namespaces for metadata locks.
  
  Introduced concept of lock duration. In addition to locks with
  transaction duration which work in the way which is similar to
  how locks worked before (i.e. they are released at the end of
  transaction), locks with statement and explicit duration were
  introduced.
  Locks with statement duration are automatically released at the
  end of statement. Locks with explicit duration require explicit
  release and obsolete concept of transactional sentinel.
  
  * Changed MDL_request and MDL_ticket classes to support notion
    of duration.
  * Changed MDL_context to keep locks with different duration in
    different lists. Changed code handling ticket list to take
    this into account.
  * Changed methods responsible for releasing locks to take into
    account duration of tickets. Particularly public
    MDL_context::release_lock() method now only can release
    tickets with explicit duration (there is still internal
    method which allows to specify duration). To release locks
    with statement or transaction duration one have to use
    release_statement/transactional_locks() methods.
  * Concept of savepoint for MDL subsystem now has to take into
    account locks with statement duration. Consequently
    MDL_savepoint class was introduced and methods working with
    savepoints were updated accordingly.
  * Added methods which allow to set duration for one or all
    locks in the context.
sql/mysqld.cc:
  Removed global mutex and condition variables which were used
  by old implementation of GRL.
  Also we no longer need to initialize Events::LOCK_event_metadata
  mutex as it was replaced with metadata locks on events.
sql/mysqld.h:
  Removed global variable, mutex and condition variables which
  were used by old implementation of GRL.
sql/rpl_rli.cc:
  When slave thread closes tables which were open for handling
  of RBR events ensure that it releases global IX lock which
  was acquired as protection against GRL.
sql/sp.cc:
  Adjusted code to the new signature of lock_object/routine_name(),
  to the fact that one now needs specify duration of lock when
  initializing MDL_request and to the fact that savepoints for MDL
  subsystem are now represented by MDL_savepoint class.
sql/sp_head.cc:
  Ensure that statements in stored procedures release statement
  metadata locks and thus release their protectiong against GRL
  in proper moment in time.
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request.
sql/sql_admin.cc:
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request.
sql/sql_base.cc:
  - Implemented support for new approach to acquiring protection
    against global read lock. We no longer acquire such protection
    explicitly on the basis of statement flags. Instead we always
    rely on code which is responsible for acquiring metadata locks
    on object to be changed acquiring this protection. This is
    achieved by acquiring global IX metadata lock with statement
    duration. Code doing this also responsible for checking that
    current connection has no active GRL by calling an
    Global_read_lock::can_acquire_protection() method.
    Changed code in open_table() and lock_table_names()
    accordingly.
    Note that as result of this change DDL and DML on temporary
    tables is always compatible with GRL (before it was
    incompatible in some cases and compatible in other cases).
  - To speed-up code acquiring protection against GRL introduced
    m_has_protection_against_grl member in Open_table_context
    class. It indicates that protection was already acquired
    sometime during open_tables() execution and new attempts
    can be skipped.
  - Thanks to new GRL implementation calls to broadcast_refresh()
    became unnecessary and were removed.
  - Adjusted code to the fact that one now needs specify duration
    of lock when initializing MDL_request and to the fact that
    savepoints for MDL subsystem are now represented by
    MDL_savepoint class.
sql/sql_base.h:
  Adjusted code to the fact that savepoints for MDL subsystem are
  now represented by MDL_savepoint class.
  Also introduced Open_table_context::m_has_protection_against_grl
  member which allows to avoid acquiring protection against GRL
  while opening tables if such protection was already acquired.
sql/sql_class.cc:
  Changed THD::leave_locked_tables_mode() after transactional
  sentinel for metadata locks was obsoleted by introduction of
  locks with explicit duration.
sql/sql_class.h:
  - Adjusted code to the fact that savepoints for MDL subsystem
    are now represented by MDL_savepoint class.
  - Changed Global_read_lock class according to changes in
    global read lock implementation:
    * wait_if_global_read_lock and start_waiting_global_read_lock
      are now gone. Instead code needing protection against GRL
      has to acquire global IX metadata lock with statement
      duration itself. To help it new can_acquire_protection()
      was introduced. Also as result of the above change
      m_protection_count member is gone too.
    * Added m_mdl_blocks_commits_lock member to store metadata
      lock blocking commits.
    * Adjusted code to the fact that concept of transactional
      sentinel was obsoleted by concept of lock duration.
  - Removed CF_PROTECT_AGAINST_GRL flag as it is no longer
    necessary. New GRL implementation acquires protection
    against global read lock automagically when statement
    acquires metadata locks on tables or other objects it
    is going to change.
sql/sql_db.cc:
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request.
sql/sql_handler.cc:
  Removed call to broadcast_refresh() function. It is no longer
  needed with new GRL implementation.
  Adjusted code after introducing duration concept for metadata
  locks. Particularly to the fact transactional sentinel was
  replaced with explicit duration.
sql/sql_handler.h:
  Renamed mysql_ha_move_tickets_after_trans_sentinel() to
  mysql_ha_set_explicit_lock_duration() after transactional
  sentinel was obsoleted by locks with explicit duration.
sql/sql_insert.cc:
  Adjusted code handling delaying inserts after switching to
  new GRL implementation. Now connection thread initiating
  delayed insert has to acquire global IX lock in addition
  to metadata lock on table being inserted into. This IX lock
  protects against GRL and similarly to SW lock on table being
  inserted into has to be passed to handler thread in order to
  avoid deadlocks.
sql/sql_lex.cc:
  LEX::protect_against_global_read_lock member is no longer
  necessary since protection against GRL is automatically
  taken by code acquiring metadata locks/opening tables.
sql/sql_lex.h:
  LEX::protect_against_global_read_lock member is no longer
  necessary since protection against GRL is automatically
  taken by code acquiring metadata locks/opening tables.
sql/sql_parse.cc:
  - Implemented support for new approach to acquiring protection
    against global read lock. We no longer acquire such protection
    explicitly on the basis of statement flags. Instead we always
    rely on code which is responsible for acquiring metadata locks
    on object to be changed acquiring this protection. This is
    achieved by acquiring global IX metadata lock with statement
    duration. This lock is automatically released at the end of
    statement execution.
  - Changed implementation of CREATE/DROP PROCEDURE/FUNCTION not
    to release metadata locks and thus protection against of GRL
    in the middle of statement execution.
  - Adjusted code to the fact that one now needs specify duration
    of lock when initializing MDL_request and to the fact that
    savepoints for MDL subsystem are now represented by
    MDL_savepoint class.
sql/sql_prepare.cc:
  Adjusted code to the to the fact that savepoints for MDL
  subsystem are now represented by MDL_savepoint class.
sql/sql_rename.cc:
  With new GRL implementation there is no need to explicitly
  acquire protection against GRL before renaming tables.
  This happens automatically in code which acquires metadata
  locks on tables being renamed.
sql/sql_show.cc:
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request and to the fact that
  savepoints for MDL subsystem are now represented by
  MDL_savepoint class.
sql/sql_table.cc:
  - With new GRL implementation there is no need to explicitly
    acquire protection against GRL before dropping tables.
    This happens automatically in code which acquires metadata
    locks on tables being dropped.
  - Changed mysql_alter_table() not to release lock on new table
    name explicitly and to rely on automatic release of locks
    at the end of statement instead. This was necessary since
    now MDL_context::release_lock() is supported only for locks
    for explicit duration.
sql/sql_trigger.cc:
  With new GRL implementation there is no need to explicitly
  acquire protection against GRL before changing table triggers.
  This happens automatically in code which acquires metadata
  locks on tables which triggers are to be changed.
sql/sql_update.cc:
  Fix bug exposed by GRL testing. During prepare phase acquire
  only S metadata locks instead of SW locks to keep prepare of
  multi-UPDATE compatible with concurrent LOCK TABLES WRITE
  and global read lock.
sql/sql_view.cc:
  With new GRL implementation there is no need to explicitly
  acquire protection against GRL before creating view.
  This happens automatically in code which acquires metadata
  lock on view to be created.
sql/sql_yacc.yy:
  LEX::protect_against_global_read_lock member is no longer
  necessary since protection against GRL is automatically
  taken by code acquiring metadata locks/opening tables.
sql/table.cc:
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request.
sql/table.h:
  Adjusted code to the fact that one now needs specify duration
  of lock when initializing MDL_request.
sql/transaction.cc:
  Replaced custom implementation of global read lock with
  one based on metadata locks. Consequently when doing
  commit instead of calling method of Global_read_lock
  class to acquire protection against GRL we simply acquire
  IX in COMMIT namespace.
  Also adjusted code to the fact that MDL savepoint is now
  represented by MDL_savepoint class.
2010-11-11 20:11:05 +03:00

1282 lines
45 KiB
C++
Raw Blame History

/* Copyright (C) 2000-2003 MySQL AB, 2008-2009 Sun Microsystems, Inc
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */
#include "sql_priv.h"
#include "unireg.h" // HAVE_*
#include "rpl_mi.h"
#include "rpl_rli.h"
#include "sql_base.h" // close_thread_tables
#include <my_dir.h> // For MY_STAT
#include "sql_repl.h" // For check_binlog_magic
#include "log_event.h" // Format_description_log_event, Log_event,
// FORMAT_DESCRIPTION_LOG_EVENT, ROTATE_EVENT,
// PREFIX_SQL_LOAD
#include "rpl_utility.h"
#include "transaction.h"
#include "sql_parse.h" // end_trans, ROLLBACK
static int count_relay_log_space(Relay_log_info* rli);
// Defined in slave.cc
int init_intvar_from_file(int* var, IO_CACHE* f, int default_val);
int init_strvar_from_file(char *var, int max_size, IO_CACHE *f,
const char *default_val);
Relay_log_info::Relay_log_info(bool is_slave_recovery)
:Slave_reporting_capability("SQL"),
no_storage(FALSE), replicate_same_server_id(::replicate_same_server_id),
info_fd(-1), cur_log_fd(-1), relay_log(&sync_relaylog_period),
sync_counter(0), is_relay_log_recovery(is_slave_recovery),
save_temporary_tables(0), cur_log_old_open_count(0), group_relay_log_pos(0),
event_relay_log_pos(0),
#if HAVE_purify
is_fake(FALSE),
#endif
group_master_log_pos(0), log_space_total(0), ignore_log_space_limit(0),
last_master_timestamp(0), slave_skip_counter(0),
abort_pos_wait(0), slave_run_id(0), sql_thd(0),
inited(0), abort_slave(0), slave_running(0), until_condition(UNTIL_NONE),
until_log_pos(0), retried_trans(0),
tables_to_lock(0), tables_to_lock_count(0),
last_event_start_time(0), m_flags(0)
{
DBUG_ENTER("Relay_log_info::Relay_log_info");
group_relay_log_name[0]= event_relay_log_name[0]=
group_master_log_name[0]= 0;
until_log_name[0]= ign_master_log_name_end[0]= 0;
bzero((char*) &info_file, sizeof(info_file));
bzero((char*) &cache_buf, sizeof(cache_buf));
cached_charset_invalidate();
mysql_mutex_init(key_relay_log_info_run_lock, &run_lock, MY_MUTEX_INIT_FAST);
mysql_mutex_init(key_relay_log_info_data_lock,
&data_lock, MY_MUTEX_INIT_FAST);
mysql_mutex_init(key_relay_log_info_log_space_lock,
&log_space_lock, MY_MUTEX_INIT_FAST);
mysql_cond_init(key_relay_log_info_data_cond, &data_cond, NULL);
mysql_cond_init(key_relay_log_info_start_cond, &start_cond, NULL);
mysql_cond_init(key_relay_log_info_stop_cond, &stop_cond, NULL);
mysql_cond_init(key_relay_log_info_log_space_cond, &log_space_cond, NULL);
relay_log.init_pthread_objects();
DBUG_VOID_RETURN;
}
Relay_log_info::~Relay_log_info()
{
DBUG_ENTER("Relay_log_info::~Relay_log_info");
mysql_mutex_destroy(&run_lock);
mysql_mutex_destroy(&data_lock);
mysql_mutex_destroy(&log_space_lock);
mysql_cond_destroy(&data_cond);
mysql_cond_destroy(&start_cond);
mysql_cond_destroy(&stop_cond);
mysql_cond_destroy(&log_space_cond);
relay_log.cleanup();
DBUG_VOID_RETURN;
}
int init_relay_log_info(Relay_log_info* rli,
const char* info_fname)
{
char fname[FN_REFLEN+128];
int info_fd;
const char* msg = 0;
int error = 0;
DBUG_ENTER("init_relay_log_info");
DBUG_ASSERT(!rli->no_storage); // Don't init if there is no storage
if (rli->inited) // Set if this function called
DBUG_RETURN(0);
fn_format(fname, info_fname, mysql_data_home, "", 4+32);
mysql_mutex_lock(&rli->data_lock);
info_fd = rli->info_fd;
rli->cur_log_fd = -1;
rli->slave_skip_counter=0;
rli->abort_pos_wait=0;
rli->log_space_limit= relay_log_space_limit;
rli->log_space_total= 0;
rli->tables_to_lock= 0;
rli->tables_to_lock_count= 0;
char pattern[FN_REFLEN];
(void) my_realpath(pattern, slave_load_tmpdir, 0);
if (fn_format(pattern, PREFIX_SQL_LOAD, pattern, "",
MY_SAFE_PATH | MY_RETURN_REAL_PATH) == NullS)
{
mysql_mutex_unlock(&rli->data_lock);
sql_print_error("Unable to use slave's temporary directory %s",
slave_load_tmpdir);
DBUG_RETURN(1);
}
unpack_filename(rli->slave_patternload_file, pattern);
rli->slave_patternload_file_size= strlen(rli->slave_patternload_file);
/*
The relay log will now be opened, as a SEQ_READ_APPEND IO_CACHE.
Note that the I/O thread flushes it to disk after writing every
event, in flush_master_info(mi, 1, ?).
*/
/*
For the maximum log size, we choose max_relay_log_size if it is
non-zero, max_binlog_size otherwise. If later the user does SET
GLOBAL on one of these variables, fix_max_binlog_size and
fix_max_relay_log_size will reconsider the choice (for example
if the user changes max_relay_log_size to zero, we have to
switch to using max_binlog_size for the relay log) and update
rli->relay_log.max_size (and mysql_bin_log.max_size).
*/
{
/* Reports an error and returns, if the --relay-log's path
is a directory.*/
if (opt_relay_logname &&
opt_relay_logname[strlen(opt_relay_logname) - 1] == FN_LIBCHAR)
{
mysql_mutex_unlock(&rli->data_lock);
sql_print_error("Path '%s' is a directory name, please specify \
a file name for --relay-log option", opt_relay_logname);
DBUG_RETURN(1);
}
/* Reports an error and returns, if the --relay-log-index's path
is a directory.*/
if (opt_relaylog_index_name &&
opt_relaylog_index_name[strlen(opt_relaylog_index_name) - 1]
== FN_LIBCHAR)
{
mysql_mutex_unlock(&rli->data_lock);
sql_print_error("Path '%s' is a directory name, please specify \
a file name for --relay-log-index option", opt_relaylog_index_name);
DBUG_RETURN(1);
}
char buf[FN_REFLEN];
const char *ln;
static bool name_warning_sent= 0;
ln= rli->relay_log.generate_name(opt_relay_logname, "-relay-bin",
1, buf);
/* We send the warning only at startup, not after every RESET SLAVE */
if (!opt_relay_logname && !opt_relaylog_index_name && !name_warning_sent)
{
/*
User didn't give us info to name the relay log index file.
Picking `hostname`-relay-bin.index like we do, causes replication to
fail if this slave's hostname is changed later. So, we would like to
instead require a name. But as we don't want to break many existing
setups, we only give warning, not error.
*/
sql_print_warning("Neither --relay-log nor --relay-log-index were used;"
" so replication "
"may break when this MySQL server acts as a "
"slave and has his hostname changed!! Please "
"use '--relay-log=%s' to avoid this problem.", ln);
name_warning_sent= 1;
}
/*
note, that if open() fails, we'll still have index file open
but a destructor will take care of that
*/
if (rli->relay_log.open_index_file(opt_relaylog_index_name, ln, TRUE) ||
rli->relay_log.open(ln, LOG_BIN, 0, SEQ_READ_APPEND, 0,
(max_relay_log_size ? max_relay_log_size :
max_binlog_size), 1, TRUE))
{
mysql_mutex_unlock(&rli->data_lock);
sql_print_error("Failed in open_log() called from init_relay_log_info()");
DBUG_RETURN(1);
}
rli->relay_log.is_relay_log= TRUE;
}
/* if file does not exist */
if (access(fname,F_OK))
{
/*
If someone removed the file from underneath our feet, just close
the old descriptor and re-create the old file
*/
if (info_fd >= 0)
mysql_file_close(info_fd, MYF(MY_WME));
if ((info_fd= mysql_file_open(key_file_relay_log_info,
fname, O_CREAT|O_RDWR|O_BINARY, MYF(MY_WME))) < 0)
{
sql_print_error("Failed to create a new relay log info file (\
file '%s', errno %d)", fname, my_errno);
msg= current_thd->stmt_da->message();
goto err;
}
if (init_io_cache(&rli->info_file, info_fd, IO_SIZE*2, READ_CACHE, 0L,0,
MYF(MY_WME)))
{
sql_print_error("Failed to create a cache on relay log info file '%s'",
fname);
msg= current_thd->stmt_da->message();
goto err;
}
/* Init relay log with first entry in the relay index file */
if (init_relay_log_pos(rli,NullS,BIN_LOG_HEADER_SIZE,0 /* no data lock */,
&msg, 0))
{
sql_print_error("Failed to open the relay log 'FIRST' (relay_log_pos 4)");
goto err;
}
rli->group_master_log_name[0]= 0;
rli->group_master_log_pos= 0;
rli->info_fd= info_fd;
}
else // file exists
{
if (info_fd >= 0)
reinit_io_cache(&rli->info_file, READ_CACHE, 0L,0,0);
else
{
int error=0;
if ((info_fd= mysql_file_open(key_file_relay_log_info,
fname, O_RDWR|O_BINARY, MYF(MY_WME))) < 0)
{
sql_print_error("\
Failed to open the existing relay log info file '%s' (errno %d)",
fname, my_errno);
error= 1;
}
else if (init_io_cache(&rli->info_file, info_fd,
IO_SIZE*2, READ_CACHE, 0L, 0, MYF(MY_WME)))
{
sql_print_error("Failed to create a cache on relay log info file '%s'",
fname);
error= 1;
}
if (error)
{
if (info_fd >= 0)
mysql_file_close(info_fd, MYF(0));
rli->info_fd= -1;
rli->relay_log.close(LOG_CLOSE_INDEX | LOG_CLOSE_STOP_EVENT);
mysql_mutex_unlock(&rli->data_lock);
DBUG_RETURN(1);
}
}
rli->info_fd = info_fd;
int relay_log_pos, master_log_pos;
if (init_strvar_from_file(rli->group_relay_log_name,
sizeof(rli->group_relay_log_name),
&rli->info_file, "") ||
init_intvar_from_file(&relay_log_pos,
&rli->info_file, BIN_LOG_HEADER_SIZE) ||
init_strvar_from_file(rli->group_master_log_name,
sizeof(rli->group_master_log_name),
&rli->info_file, "") ||
init_intvar_from_file(&master_log_pos, &rli->info_file, 0))
{
msg="Error reading slave log configuration";
goto err;
}
strmake(rli->event_relay_log_name,rli->group_relay_log_name,
sizeof(rli->event_relay_log_name)-1);
rli->group_relay_log_pos= rli->event_relay_log_pos= relay_log_pos;
rli->group_master_log_pos= master_log_pos;
if (rli->is_relay_log_recovery && init_recovery(rli->mi, &msg))
goto err;
if (init_relay_log_pos(rli,
rli->group_relay_log_name,
rli->group_relay_log_pos,
0 /* no data lock*/,
&msg, 0))
{
char llbuf[22];
sql_print_error("Failed to open the relay log '%s' (relay_log_pos %s)",
rli->group_relay_log_name,
llstr(rli->group_relay_log_pos, llbuf));
goto err;
}
}
#ifndef DBUG_OFF
{
char llbuf1[22], llbuf2[22];
DBUG_PRINT("info", ("my_b_tell(rli->cur_log)=%s rli->event_relay_log_pos=%s",
llstr(my_b_tell(rli->cur_log),llbuf1),
llstr(rli->event_relay_log_pos,llbuf2)));
DBUG_ASSERT(rli->event_relay_log_pos >= BIN_LOG_HEADER_SIZE);
DBUG_ASSERT(my_b_tell(rli->cur_log) == rli->event_relay_log_pos);
}
#endif
/*
Now change the cache from READ to WRITE - must do this
before flush_relay_log_info
*/
reinit_io_cache(&rli->info_file, WRITE_CACHE,0L,0,1);
if ((error= flush_relay_log_info(rli)))
{
msg= "Failed to flush relay log info file";
goto err;
}
if (count_relay_log_space(rli))
{
msg="Error counting relay log space";
goto err;
}
rli->inited= 1;
mysql_mutex_unlock(&rli->data_lock);
DBUG_RETURN(error);
err:
sql_print_error("%s", msg);
end_io_cache(&rli->info_file);
if (info_fd >= 0)
mysql_file_close(info_fd, MYF(0));
rli->info_fd= -1;
rli->relay_log.close(LOG_CLOSE_INDEX | LOG_CLOSE_STOP_EVENT);
mysql_mutex_unlock(&rli->data_lock);
DBUG_RETURN(1);
}
static inline int add_relay_log(Relay_log_info* rli,LOG_INFO* linfo)
{
MY_STAT s;
DBUG_ENTER("add_relay_log");
if (!mysql_file_stat(key_file_binlog,
linfo->log_file_name, &s, MYF(0)))
{
sql_print_error("log %s listed in the index, but failed to stat",
linfo->log_file_name);
DBUG_RETURN(1);
}
rli->log_space_total += s.st_size;
#ifndef DBUG_OFF
char buf[22];
DBUG_PRINT("info",("log_space_total: %s", llstr(rli->log_space_total,buf)));
#endif
DBUG_RETURN(0);
}
static int count_relay_log_space(Relay_log_info* rli)
{
LOG_INFO linfo;
DBUG_ENTER("count_relay_log_space");
rli->log_space_total= 0;
if (rli->relay_log.find_log_pos(&linfo, NullS, 1))
{
sql_print_error("Could not find first log while counting relay log space");
DBUG_RETURN(1);
}
do
{
if (add_relay_log(rli,&linfo))
DBUG_RETURN(1);
} while (!rli->relay_log.find_next_log(&linfo, 1));
/*
As we have counted everything, including what may have written in a
preceding write, we must reset bytes_written, or we may count some space
twice.
*/
rli->relay_log.reset_bytes_written();
DBUG_RETURN(0);
}
/*
Reset UNTIL condition for Relay_log_info
SYNOPSYS
clear_until_condition()
rli - Relay_log_info structure where UNTIL condition should be reset
*/
void Relay_log_info::clear_until_condition()
{
DBUG_ENTER("clear_until_condition");
until_condition= Relay_log_info::UNTIL_NONE;
until_log_name[0]= 0;
until_log_pos= 0;
DBUG_VOID_RETURN;
}
/*
Open the given relay log
SYNOPSIS
init_relay_log_pos()
rli Relay information (will be initialized)
log Name of relay log file to read from. NULL = First log
pos Position in relay log file
need_data_lock Set to 1 if this functions should do mutex locks
errmsg Store pointer to error message here
look_for_description_event
1 if we should look for such an event. We only need
this when the SQL thread starts and opens an existing
relay log and has to execute it (possibly from an
offset >4); then we need to read the first event of
the relay log to be able to parse the events we have
to execute.
DESCRIPTION
- Close old open relay log files.
- If we are using the same relay log as the running IO-thread, then set
rli->cur_log to point to the same IO_CACHE entry.
- If not, open the 'log' binary file.
TODO
- check proper initialization of group_master_log_name/group_master_log_pos
RETURN VALUES
0 ok
1 error. errmsg is set to point to the error message
*/
int init_relay_log_pos(Relay_log_info* rli,const char* log,
ulonglong pos, bool need_data_lock,
const char** errmsg,
bool look_for_description_event)
{
DBUG_ENTER("init_relay_log_pos");
DBUG_PRINT("info", ("pos: %lu", (ulong) pos));
*errmsg=0;
mysql_mutex_t *log_lock= rli->relay_log.get_log_lock();
if (need_data_lock)
mysql_mutex_lock(&rli->data_lock);
/*
Slave threads are not the only users of init_relay_log_pos(). CHANGE MASTER
is, too, and init_slave() too; these 2 functions allocate a description
event in init_relay_log_pos, which is not freed by the terminating SQL slave
thread as that thread is not started by these functions. So we have to free
the description_event here, in case, so that there is no memory leak in
running, say, CHANGE MASTER.
*/
delete rli->relay_log.description_event_for_exec;
/*
By default the relay log is in binlog format 3 (4.0).
Even if format is 4, this will work enough to read the first event
(Format_desc) (remember that format 4 is just lenghtened compared to format
3; format 3 is a prefix of format 4).
*/
rli->relay_log.description_event_for_exec= new
Format_description_log_event(3);
mysql_mutex_lock(log_lock);
/* Close log file and free buffers if it's already open */
if (rli->cur_log_fd >= 0)
{
end_io_cache(&rli->cache_buf);
mysql_file_close(rli->cur_log_fd, MYF(MY_WME));
rli->cur_log_fd = -1;
}
rli->group_relay_log_pos = rli->event_relay_log_pos = pos;
/*
Test to see if the previous run was with the skip of purging
If yes, we do not purge when we restart
*/
if (rli->relay_log.find_log_pos(&rli->linfo, NullS, 1))
{
*errmsg="Could not find first log during relay log initialization";
goto err;
}
if (log && rli->relay_log.find_log_pos(&rli->linfo, log, 1))
{
*errmsg="Could not find target log during relay log initialization";
goto err;
}
strmake(rli->group_relay_log_name,rli->linfo.log_file_name,
sizeof(rli->group_relay_log_name)-1);
strmake(rli->event_relay_log_name,rli->linfo.log_file_name,
sizeof(rli->event_relay_log_name)-1);
if (rli->relay_log.is_active(rli->linfo.log_file_name))
{
/*
The IO thread is using this log file.
In this case, we will use the same IO_CACHE pointer to
read data as the IO thread is using to write data.
*/
my_b_seek((rli->cur_log=rli->relay_log.get_log_file()), (off_t)0);
if (check_binlog_magic(rli->cur_log,errmsg))
goto err;
rli->cur_log_old_open_count=rli->relay_log.get_open_count();
}
else
{
/*
Open the relay log and set rli->cur_log to point at this one
*/
if ((rli->cur_log_fd=open_binlog(&rli->cache_buf,
rli->linfo.log_file_name,errmsg)) < 0)
goto err;
rli->cur_log = &rli->cache_buf;
}
/*
In all cases, check_binlog_magic() has been called so we're at offset 4 for
sure.
*/
if (pos > BIN_LOG_HEADER_SIZE) /* If pos<=4, we stay at 4 */
{
Log_event* ev;
while (look_for_description_event)
{
/*
Read the possible Format_description_log_event; if position
was 4, no need, it will be read naturally.
*/
DBUG_PRINT("info",("looking for a Format_description_log_event"));
if (my_b_tell(rli->cur_log) >= pos)
break;
/*
Because of we have rli->data_lock and log_lock, we can safely read an
event
*/
if (!(ev=Log_event::read_log_event(rli->cur_log,0,
rli->relay_log.description_event_for_exec)))
{
DBUG_PRINT("info",("could not read event, rli->cur_log->error=%d",
rli->cur_log->error));
if (rli->cur_log->error) /* not EOF */
{
*errmsg= "I/O error reading event at position 4";
goto err;
}
break;
}
else if (ev->get_type_code() == FORMAT_DESCRIPTION_EVENT)
{
DBUG_PRINT("info",("found Format_description_log_event"));
delete rli->relay_log.description_event_for_exec;
rli->relay_log.description_event_for_exec= (Format_description_log_event*) ev;
/*
As ev was returned by read_log_event, it has passed is_valid(), so
my_malloc() in ctor worked, no need to check again.
*/
/*
Ok, we found a Format_description event. But it is not sure that this
describes the whole relay log; indeed, one can have this sequence
(starting from position 4):
Format_desc (of slave)
Rotate (of master)
Format_desc (of master)
So the Format_desc which really describes the rest of the relay log
is the 3rd event (it can't be further than that, because we rotate
the relay log when we queue a Rotate event from the master).
But what describes the Rotate is the first Format_desc.
So what we do is:
go on searching for Format_description events, until you exceed the
position (argument 'pos') or until you find another event than Rotate
or Format_desc.
*/
}
else
{
DBUG_PRINT("info",("found event of another type=%d",
ev->get_type_code()));
look_for_description_event= (ev->get_type_code() == ROTATE_EVENT);
delete ev;
}
}
my_b_seek(rli->cur_log,(off_t)pos);
#ifndef DBUG_OFF
{
char llbuf1[22], llbuf2[22];
DBUG_PRINT("info", ("my_b_tell(rli->cur_log)=%s rli->event_relay_log_pos=%s",
llstr(my_b_tell(rli->cur_log),llbuf1),
llstr(rli->event_relay_log_pos,llbuf2)));
}
#endif
}
err:
/*
If we don't purge, we can't honour relay_log_space_limit ;
silently discard it
*/
if (!relay_log_purge)
rli->log_space_limit= 0;
mysql_cond_broadcast(&rli->data_cond);
mysql_mutex_unlock(log_lock);
if (need_data_lock)
mysql_mutex_unlock(&rli->data_lock);
if (!rli->relay_log.description_event_for_exec->is_valid() && !*errmsg)
*errmsg= "Invalid Format_description log event; could be out of memory";
DBUG_RETURN ((*errmsg) ? 1 : 0);
}
/*
Waits until the SQL thread reaches (has executed up to) the
log/position or timed out.
SYNOPSIS
wait_for_pos()
thd client thread that sent SELECT MASTER_POS_WAIT
log_name log name to wait for
log_pos position to wait for
timeout timeout in seconds before giving up waiting
NOTES
timeout is longlong whereas it should be ulong ; but this is
to catch if the user submitted a negative timeout.
RETURN VALUES
-2 improper arguments (log_pos<0)
or slave not running, or master info changed
during the function's execution,
or client thread killed. -2 is translated to NULL by caller
-1 timed out
>=0 number of log events the function had to wait
before reaching the desired log/position
*/
int Relay_log_info::wait_for_pos(THD* thd, String* log_name,
longlong log_pos,
longlong timeout)
{
int event_count = 0;
ulong init_abort_pos_wait;
int error=0;
struct timespec abstime; // for timeout checking
const char *msg;
DBUG_ENTER("Relay_log_info::wait_for_pos");
if (!inited)
DBUG_RETURN(-2);
DBUG_PRINT("enter",("log_name: '%s' log_pos: %lu timeout: %lu",
log_name->c_ptr(), (ulong) log_pos, (ulong) timeout));
set_timespec(abstime,timeout);
mysql_mutex_lock(&data_lock);
msg= thd->enter_cond(&data_cond, &data_lock,
"Waiting for the slave SQL thread to "
"advance position");
/*
This function will abort when it notices that some CHANGE MASTER or
RESET MASTER has changed the master info.
To catch this, these commands modify abort_pos_wait ; We just monitor
abort_pos_wait and see if it has changed.
Why do we have this mechanism instead of simply monitoring slave_running
in the loop (we do this too), as CHANGE MASTER/RESET SLAVE require that
the SQL thread be stopped?
This is becasue if someones does:
STOP SLAVE;CHANGE MASTER/RESET SLAVE; START SLAVE;
the change may happen very quickly and we may not notice that
slave_running briefly switches between 1/0/1.
*/
init_abort_pos_wait= abort_pos_wait;
/*
We'll need to
handle all possible log names comparisons (e.g. 999 vs 1000).
We use ulong for string->number conversion ; this is no
stronger limitation than in find_uniq_filename in sql/log.cc
*/
ulong log_name_extension;
char log_name_tmp[FN_REFLEN]; //make a char[] from String
strmake(log_name_tmp, log_name->ptr(), min(log_name->length(), FN_REFLEN-1));
char *p= fn_ext(log_name_tmp);
char *p_end;
if (!*p || log_pos<0)
{
error= -2; //means improper arguments
goto err;
}
// Convert 0-3 to 4
log_pos= max(log_pos, BIN_LOG_HEADER_SIZE);
/* p points to '.' */
log_name_extension= strtoul(++p, &p_end, 10);
/*
p_end points to the first invalid character.
If it equals to p, no digits were found, error.
If it contains '\0' it means conversion went ok.
*/
if (p_end==p || *p_end)
{
error= -2;
goto err;
}
/* The "compare and wait" main loop */
while (!thd->killed &&
init_abort_pos_wait == abort_pos_wait &&
slave_running)
{
bool pos_reached;
int cmp_result= 0;
DBUG_PRINT("info",
("init_abort_pos_wait: %ld abort_pos_wait: %ld",
init_abort_pos_wait, abort_pos_wait));
DBUG_PRINT("info",("group_master_log_name: '%s' pos: %lu",
group_master_log_name, (ulong) group_master_log_pos));
/*
group_master_log_name can be "", if we are just after a fresh
replication start or after a CHANGE MASTER TO MASTER_HOST/PORT
(before we have executed one Rotate event from the master) or
(rare) if the user is doing a weird slave setup (see next
paragraph). If group_master_log_name is "", we assume we don't
have enough info to do the comparison yet, so we just wait until
more data. In this case master_log_pos is always 0 except if
somebody (wrongly) sets this slave to be a slave of itself
without using --replicate-same-server-id (an unsupported
configuration which does nothing), then group_master_log_pos
will grow and group_master_log_name will stay "".
*/
if (*group_master_log_name)
{
char *basename= (group_master_log_name +
dirname_length(group_master_log_name));
/*
First compare the parts before the extension.
Find the dot in the master's log basename,
and protect against user's input error :
if the names do not match up to '.' included, return error
*/
char *q= (char*)(fn_ext(basename)+1);
if (strncmp(basename, log_name_tmp, (int)(q-basename)))
{
error= -2;
break;
}
// Now compare extensions.
char *q_end;
ulong group_master_log_name_extension= strtoul(q, &q_end, 10);
if (group_master_log_name_extension < log_name_extension)
cmp_result= -1 ;
else
cmp_result= (group_master_log_name_extension > log_name_extension) ? 1 : 0 ;
pos_reached= ((!cmp_result && group_master_log_pos >= (ulonglong)log_pos) ||
cmp_result > 0);
if (pos_reached || thd->killed)
break;
}
//wait for master update, with optional timeout.
DBUG_PRINT("info",("Waiting for master update"));
/*
We are going to mysql_cond_(timed)wait(); if the SQL thread stops it
will wake us up.
*/
if (timeout > 0)
{
/*
Note that mysql_cond_timedwait checks for the timeout
before for the condition ; i.e. it returns ETIMEDOUT
if the system time equals or exceeds the time specified by abstime
before the condition variable is signaled or broadcast, _or_ if
the absolute time specified by abstime has already passed at the time
of the call.
For that reason, mysql_cond_timedwait will do the "timeoutting" job
even if its condition is always immediately signaled (case of a loaded
master).
*/
error= mysql_cond_timedwait(&data_cond, &data_lock, &abstime);
}
else
mysql_cond_wait(&data_cond, &data_lock);
DBUG_PRINT("info",("Got signal of master update or timed out"));
if (error == ETIMEDOUT || error == ETIME)
{
error= -1;
break;
}
error=0;
event_count++;
DBUG_PRINT("info",("Testing if killed or SQL thread not running"));
}
err:
thd->exit_cond(msg);
DBUG_PRINT("exit",("killed: %d abort: %d slave_running: %d \
improper_arguments: %d timed_out: %d",
thd->killed_errno(),
(int) (init_abort_pos_wait != abort_pos_wait),
(int) slave_running,
(int) (error == -2),
(int) (error == -1)));
if (thd->killed || init_abort_pos_wait != abort_pos_wait ||
!slave_running)
{
error= -2;
}
DBUG_RETURN( error ? error : event_count );
}
void Relay_log_info::inc_group_relay_log_pos(ulonglong log_pos,
bool skip_lock)
{
DBUG_ENTER("Relay_log_info::inc_group_relay_log_pos");
if (!skip_lock)
mysql_mutex_lock(&data_lock);
inc_event_relay_log_pos();
group_relay_log_pos= event_relay_log_pos;
strmake(group_relay_log_name,event_relay_log_name,
sizeof(group_relay_log_name)-1);
notify_group_relay_log_name_update();
/*
If the slave does not support transactions and replicates a transaction,
users should not trust group_master_log_pos (which they can display with
SHOW SLAVE STATUS or read from relay-log.info), because to compute
group_master_log_pos the slave relies on log_pos stored in the master's
binlog, but if we are in a master's transaction these positions are always
the BEGIN's one (excepted for the COMMIT), so group_master_log_pos does
not advance as it should on the non-transactional slave (it advances by
big leaps, whereas it should advance by small leaps).
*/
/*
In 4.x we used the event's len to compute the positions here. This is
wrong if the event was 3.23/4.0 and has been converted to 5.0, because
then the event's len is not what is was in the master's binlog, so this
will make a wrong group_master_log_pos (yes it's a bug in 3.23->4.0
replication: Exec_master_log_pos is wrong). Only way to solve this is to
have the original offset of the end of the event the relay log. This is
what we do in 5.0: log_pos has become "end_log_pos" (because the real use
of log_pos in 4.0 was to compute the end_log_pos; so better to store
end_log_pos instead of begin_log_pos.
If we had not done this fix here, the problem would also have appeared
when the slave and master are 5.0 but with different event length (for
example the slave is more recent than the master and features the event
UID). It would give false MASTER_POS_WAIT, false Exec_master_log_pos in
SHOW SLAVE STATUS, and so the user would do some CHANGE MASTER using this
value which would lead to badly broken replication.
Even the relay_log_pos will be corrupted in this case, because the len is
the relay log is not "val".
With the end_log_pos solution, we avoid computations involving lengthes.
*/
DBUG_PRINT("info", ("log_pos: %lu group_master_log_pos: %lu",
(long) log_pos, (long) group_master_log_pos));
if (log_pos) // 3.23 binlogs don't have log_posx
{
group_master_log_pos= log_pos;
}
mysql_cond_broadcast(&data_cond);
if (!skip_lock)
mysql_mutex_unlock(&data_lock);
DBUG_VOID_RETURN;
}
void Relay_log_info::close_temporary_tables()
{
TABLE *table,*next;
DBUG_ENTER("Relay_log_info::close_temporary_tables");
for (table=save_temporary_tables ; table ; table=next)
{
next=table->next;
/*
Don't ask for disk deletion. For now, anyway they will be deleted when
slave restarts, but it is a better intention to not delete them.
*/
DBUG_PRINT("info", ("table: 0x%lx", (long) table));
close_temporary(table, 1, 0);
}
save_temporary_tables= 0;
slave_open_temp_tables= 0;
DBUG_VOID_RETURN;
}
/*
purge_relay_logs()
NOTES
Assumes to have a run lock on rli and that no slave thread are running.
*/
int purge_relay_logs(Relay_log_info* rli, THD *thd, bool just_reset,
const char** errmsg)
{
int error=0;
DBUG_ENTER("purge_relay_logs");
/*
Even if rli->inited==0, we still try to empty rli->master_log_* variables.
Indeed, rli->inited==0 does not imply that they already are empty.
It could be that slave's info initialization partly succeeded :
for example if relay-log.info existed but *relay-bin*.*
have been manually removed, init_relay_log_info reads the old
relay-log.info and fills rli->master_log_*, then init_relay_log_info
checks for the existence of the relay log, this fails and
init_relay_log_info leaves rli->inited to 0.
In that pathological case, rli->master_log_pos* will be properly reinited
at the next START SLAVE (as RESET SLAVE or CHANGE
MASTER, the callers of purge_relay_logs, will delete bogus *.info files
or replace them with correct files), however if the user does SHOW SLAVE
STATUS before START SLAVE, he will see old, confusing rli->master_log_*.
In other words, we reinit rli->master_log_* for SHOW SLAVE STATUS
to display fine in any case.
*/
rli->group_master_log_name[0]= 0;
rli->group_master_log_pos= 0;
if (!rli->inited)
{
DBUG_PRINT("info", ("rli->inited == 0"));
DBUG_RETURN(0);
}
DBUG_ASSERT(rli->slave_running == 0);
DBUG_ASSERT(rli->mi->slave_running == 0);
rli->slave_skip_counter=0;
mysql_mutex_lock(&rli->data_lock);
/*
we close the relay log fd possibly left open by the slave SQL thread,
to be able to delete it; the relay log fd possibly left open by the slave
I/O thread will be closed naturally in reset_logs() by the
close(LOG_CLOSE_TO_BE_OPENED) call
*/
if (rli->cur_log_fd >= 0)
{
end_io_cache(&rli->cache_buf);
mysql_file_close(rli->cur_log_fd, MYF(MY_WME));
rli->cur_log_fd= -1;
}
if (rli->relay_log.reset_logs(thd))
{
*errmsg = "Failed during log reset";
error=1;
goto err;
}
/* Save name of used relay log file */
strmake(rli->group_relay_log_name, rli->relay_log.get_log_fname(),
sizeof(rli->group_relay_log_name)-1);
strmake(rli->event_relay_log_name, rli->relay_log.get_log_fname(),
sizeof(rli->event_relay_log_name)-1);
rli->group_relay_log_pos= rli->event_relay_log_pos= BIN_LOG_HEADER_SIZE;
if (count_relay_log_space(rli))
{
*errmsg= "Error counting relay log space";
error=1;
goto err;
}
if (!just_reset)
error= init_relay_log_pos(rli, rli->group_relay_log_name,
rli->group_relay_log_pos,
0 /* do not need data lock */, errmsg, 0);
err:
#ifndef DBUG_OFF
char buf[22];
#endif
DBUG_PRINT("info",("log_space_total: %s",llstr(rli->log_space_total,buf)));
mysql_mutex_unlock(&rli->data_lock);
DBUG_RETURN(error);
}
/*
Check if condition stated in UNTIL clause of START SLAVE is reached.
SYNOPSYS
Relay_log_info::is_until_satisfied()
master_beg_pos position of the beginning of to be executed event
(not log_pos member of the event that points to the
beginning of the following event)
DESCRIPTION
Checks if UNTIL condition is reached. Uses caching result of last
comparison of current log file name and target log file name. So cached
value should be invalidated if current log file name changes
(see Relay_log_info::notify_... functions).
This caching is needed to avoid of expensive string comparisons and
strtol() conversions needed for log names comparison. We don't need to
compare them each time this function is called, we only need to do this
when current log name changes. If we have UNTIL_MASTER_POS condition we
need to do this only after Rotate_log_event::do_apply_event() (which is
rare, so caching gives real benifit), and if we have UNTIL_RELAY_POS
condition then we should invalidate cached comarison value after
inc_group_relay_log_pos() which called for each group of events (so we
have some benefit if we have something like queries that use
autoincrement or if we have transactions).
Should be called ONLY if until_condition != UNTIL_NONE !
RETURN VALUE
true - condition met or error happened (condition seems to have
bad log file name)
false - condition not met
*/
bool Relay_log_info::is_until_satisfied(THD *thd, Log_event *ev)
{
const char *log_name;
ulonglong log_pos;
DBUG_ENTER("Relay_log_info::is_until_satisfied");
DBUG_ASSERT(until_condition != UNTIL_NONE);
if (until_condition == UNTIL_MASTER_POS)
{
if (ev && ev->server_id == (uint32) ::server_id && !replicate_same_server_id)
DBUG_RETURN(FALSE);
log_name= group_master_log_name;
log_pos= (!ev)? group_master_log_pos :
((thd->variables.option_bits & OPTION_BEGIN || !ev->log_pos) ?
group_master_log_pos : ev->log_pos - ev->data_written);
}
else
{ /* until_condition == UNTIL_RELAY_POS */
log_name= group_relay_log_name;
log_pos= group_relay_log_pos;
}
#ifndef DBUG_OFF
{
char buf[32];
DBUG_PRINT("info", ("group_master_log_name='%s', group_master_log_pos=%s",
group_master_log_name, llstr(group_master_log_pos, buf)));
DBUG_PRINT("info", ("group_relay_log_name='%s', group_relay_log_pos=%s",
group_relay_log_name, llstr(group_relay_log_pos, buf)));
DBUG_PRINT("info", ("(%s) log_name='%s', log_pos=%s",
until_condition == UNTIL_MASTER_POS ? "master" : "relay",
log_name, llstr(log_pos, buf)));
DBUG_PRINT("info", ("(%s) until_log_name='%s', until_log_pos=%s",
until_condition == UNTIL_MASTER_POS ? "master" : "relay",
until_log_name, llstr(until_log_pos, buf)));
}
#endif
if (until_log_names_cmp_result == UNTIL_LOG_NAMES_CMP_UNKNOWN)
{
/*
We have no cached comparison results so we should compare log names
and cache result.
If we are after RESET SLAVE, and the SQL slave thread has not processed
any event yet, it could be that group_master_log_name is "". In that case,
just wait for more events (as there is no sensible comparison to do).
*/
if (*log_name)
{
const char *basename= log_name + dirname_length(log_name);
const char *q= (const char*)(fn_ext(basename)+1);
if (strncmp(basename, until_log_name, (int)(q-basename)) == 0)
{
/* Now compare extensions. */
char *q_end;
ulong log_name_extension= strtoul(q, &q_end, 10);
if (log_name_extension < until_log_name_extension)
until_log_names_cmp_result= UNTIL_LOG_NAMES_CMP_LESS;
else
until_log_names_cmp_result=
(log_name_extension > until_log_name_extension) ?
UNTIL_LOG_NAMES_CMP_GREATER : UNTIL_LOG_NAMES_CMP_EQUAL ;
}
else
{
/* Probably error so we aborting */
sql_print_error("Slave SQL thread is stopped because UNTIL "
"condition is bad.");
DBUG_RETURN(TRUE);
}
}
else
DBUG_RETURN(until_log_pos == 0);
}
DBUG_RETURN(((until_log_names_cmp_result == UNTIL_LOG_NAMES_CMP_EQUAL &&
log_pos >= until_log_pos) ||
until_log_names_cmp_result == UNTIL_LOG_NAMES_CMP_GREATER));
}
void Relay_log_info::cached_charset_invalidate()
{
DBUG_ENTER("Relay_log_info::cached_charset_invalidate");
/* Full of zeroes means uninitialized. */
bzero(cached_charset, sizeof(cached_charset));
DBUG_VOID_RETURN;
}
bool Relay_log_info::cached_charset_compare(char *charset) const
{
DBUG_ENTER("Relay_log_info::cached_charset_compare");
if (memcmp(cached_charset, charset, sizeof(cached_charset)))
{
memcpy(const_cast<char*>(cached_charset), charset, sizeof(cached_charset));
DBUG_RETURN(1);
}
DBUG_RETURN(0);
}
void Relay_log_info::stmt_done(my_off_t event_master_log_pos,
time_t event_creation_time)
{
#ifndef DBUG_OFF
extern uint debug_not_change_ts_if_art_event;
#endif
clear_flag(IN_STMT);
/*
If in a transaction, and if the slave supports transactions, just
inc_event_relay_log_pos(). We only have to check for OPTION_BEGIN
(not OPTION_NOT_AUTOCOMMIT) as transactions are logged with
BEGIN/COMMIT, not with SET AUTOCOMMIT= .
CAUTION: opt_using_transactions means innodb || bdb ; suppose the
master supports InnoDB and BDB, but the slave supports only BDB,
problems will arise: - suppose an InnoDB table is created on the
master, - then it will be MyISAM on the slave - but as
opt_using_transactions is true, the slave will believe he is
transactional with the MyISAM table. And problems will come when
one does START SLAVE; STOP SLAVE; START SLAVE; (the slave will
resume at BEGIN whereas there has not been any rollback). This is
the problem of using opt_using_transactions instead of a finer
"does the slave support _transactional handler used on the
master_".
More generally, we'll have problems when a query mixes a
transactional handler and MyISAM and STOP SLAVE is issued in the
middle of the "transaction". START SLAVE will resume at BEGIN
while the MyISAM table has already been updated.
*/
if ((sql_thd->variables.option_bits & OPTION_BEGIN) && opt_using_transactions)
inc_event_relay_log_pos();
else
{
inc_group_relay_log_pos(event_master_log_pos);
flush_relay_log_info(this);
/*
Note that Rotate_log_event::do_apply_event() does not call this
function, so there is no chance that a fake rotate event resets
last_master_timestamp. Note that we update without mutex
(probably ok - except in some very rare cases, only consequence
is that value may take some time to display in
Seconds_Behind_Master - not critical).
*/
#ifndef DBUG_OFF
if (!(event_creation_time == 0 && debug_not_change_ts_if_art_event > 0))
#else
if (event_creation_time != 0)
#endif
last_master_timestamp= event_creation_time;
}
}
#if !defined(MYSQL_CLIENT) && defined(HAVE_REPLICATION)
void Relay_log_info::cleanup_context(THD *thd, bool error)
{
DBUG_ENTER("Relay_log_info::cleanup_context");
DBUG_ASSERT(sql_thd == thd);
/*
1) Instances of Table_map_log_event, if ::do_apply_event() was called on them,
may have opened tables, which we cannot be sure have been closed (because
maybe the Rows_log_event have not been found or will not be, because slave
SQL thread is stopping, or relay log has a missing tail etc). So we close
all thread's tables. And so the table mappings have to be cancelled.
2) Rows_log_event::do_apply_event() may even have started statements or
transactions on them, which we need to rollback in case of error.
3) If finding a Format_description_log_event after a BEGIN, we also need
to rollback before continuing with the next events.
4) so we need this "context cleanup" function.
*/
if (error)
{
trans_rollback_stmt(thd); // if a "statement transaction"
trans_rollback(thd); // if a "real transaction"
}
m_table_map.clear_tables();
slave_close_thread_tables(thd);
if (error)
thd->mdl_context.release_transactional_locks();
clear_flag(IN_STMT);
/*
Cleanup for the flags that have been set at do_apply_event.
*/
thd->variables.option_bits&= ~OPTION_NO_FOREIGN_KEY_CHECKS;
thd->variables.option_bits&= ~OPTION_RELAXED_UNIQUE_CHECKS;
DBUG_VOID_RETURN;
}
void Relay_log_info::clear_tables_to_lock()
{
while (tables_to_lock)
{
uchar* to_free= reinterpret_cast<uchar*>(tables_to_lock);
if (tables_to_lock->m_tabledef_valid)
{
tables_to_lock->m_tabledef.table_def::~table_def();
tables_to_lock->m_tabledef_valid= FALSE;
}
tables_to_lock=
static_cast<RPL_TABLE_LIST*>(tables_to_lock->next_global);
tables_to_lock_count--;
my_free(to_free);
}
DBUG_ASSERT(tables_to_lock == NULL && tables_to_lock_count == 0);
}
void Relay_log_info::slave_close_thread_tables(THD *thd)
{
thd->stmt_da->can_overwrite_status= TRUE;
thd->is_error() ? trans_rollback_stmt(thd) : trans_commit_stmt(thd);
thd->stmt_da->can_overwrite_status= FALSE;
close_thread_tables(thd);
/*
- If inside a multi-statement transaction,
defer the release of metadata locks until the current
transaction is either committed or rolled back. This prevents
other statements from modifying the table for the entire
duration of this transaction. This provides commit ordering
and guarantees serializability across multiple transactions.
- If in autocommit mode, or outside a transactional context,
automatically release metadata locks of the current statement.
*/
if (! thd->in_multi_stmt_transaction_mode())
thd->mdl_context.release_transactional_locks();
else
thd->mdl_context.release_statement_locks();
clear_tables_to_lock();
}
#endif
Reference in a new issue View git blame Copy permalink