mirror of
https://github.com/MariaDB/server.git
synced 2026-05-06 15:15:34 +02:00
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
sql/handler.cc:
Moved logic for system / log tables in the SQL layer.
sql/handler.h:
Moved logic for system / log tables in the SQL layer.
sql/lock.cc:
Revised locking of log tables
sql/log.cc:
Major cleanup: changed how log tables are locked / written to.
sql/log.h:
Major cleanup: changed how log tables are locked / written to.
sql/mysql_priv.h:
performance schema helpers
sql/slave.cc:
open_ltable() lock flags
sql/sp.cc:
open_ltable() lock flags
sql/sql_acl.cc:
open_ltable() lock flags
sql/sql_class.h:
performance schema helpers
sql/sql_delete.cc:
log tables cleanup in TRUNCATE
sql/sql_error.cc:
Internal handlers can also intercept warnings
sql/sql_insert.cc:
open_ltable() lock flags
sql/sql_parse.cc:
performance schema helpers
sql/sql_plugin.cc:
open_ltable() lock flags
sql/sql_rename.cc:
log tables cleanup in RENAME
sql/sql_servers.cc:
open_ltable() lock flags
sql/sql_show.cc:
Move INFORMATION_SCHEMA_NAME to table.cc
sql/sql_table.cc:
log tables cleanup (admin operations, ALTER TABLE)
sql/sql_udf.cc:
open_ltable() lock flags
sql/table.cc:
Implemented TABLE_CATEGORY.
sql/share/errmsg.txt:
Changed the wording and name of ER_CANT_READ_LOCK_LOG_TABLE
sql/table.h:
Implemented TABLE_CATEGORY.
storage/csv/ha_tina.cc:
Moved logic for system / log tables in the SQL layer.
storage/csv/ha_tina.h:
Moved logic for system / log tables in the SQL layer.
storage/myisam/ha_myisam.cc:
Moved logic for system / log tables in the SQL layer.
storage/myisam/ha_myisam.h:
Moved logic for system / log tables in the SQL layer.
client/mysqldump.c:
Don't lock tables in the ignore list.
Don't issue empty LOCK TABLES queries.
sql/sql_base.cc:
log tables cleanup
performance schema helpers
mysql-test/r/ps.result:
Adjust test results
mysql-test/r/show_check.result:
Adjust test results
mysql-test/r/status.result:
Adjust test results
mysql-test/t/log_state.test:
Added tests for Bug#29129
mysql-test/t/ps.test:
Make the test output deterministic
mysql-test/t/show_check.test:
Make the test output deterministic
mysql-test/r/log_state.result:
Changed the default location of the log output to LOG_FILE,
for backward compatibility with MySQL 5.0
---
Adjust test results
mysql-test/r/log_tables.result:
cleanup for -ps-protocol
mysql-test/t/log_tables.test:
cleanup for -ps-protocol
sql/set_var.cc:
Changed the default location of the log output to LOG_FILE,
for backward compatibility with MySQL 5.0
---
log tables cleanup
This commit is contained in:
parent
b0fcdce317
commit
4462578a1c
39 changed files with 1490 additions and 993 deletions
|
|
@ -1516,34 +1516,6 @@ THD *handler::ha_thd(void) const
|
|||
}
|
||||
|
||||
|
||||
bool handler::check_if_log_table_locking_is_allowed(uint sql_command,
|
||||
ulong type, TABLE *table)
|
||||
{
|
||||
/*
|
||||
Deny locking of the log tables, which is incompatible with
|
||||
concurrent insert. The routine is not called if the table is
|
||||
being locked from a logger THD (general_log_thd or slow_log_thd)
|
||||
or from a privileged thread (see log.cc for details)
|
||||
*/
|
||||
if (table->s->log_table &&
|
||||
sql_command != SQLCOM_TRUNCATE &&
|
||||
sql_command != SQLCOM_ALTER_TABLE &&
|
||||
!(sql_command == SQLCOM_FLUSH &&
|
||||
type & REFRESH_LOG) &&
|
||||
(table->reginfo.lock_type >= TL_READ_NO_INSERT))
|
||||
{
|
||||
/*
|
||||
The check >= TL_READ_NO_INSERT denies all write locks
|
||||
plus the only read lock (TL_READ_NO_INSERT itself)
|
||||
*/
|
||||
table->reginfo.lock_type == TL_READ_NO_INSERT ?
|
||||
my_error(ER_CANT_READ_LOCK_LOG_TABLE, MYF(0)) :
|
||||
my_error(ER_CANT_WRITE_LOCK_LOG_TABLE, MYF(0));
|
||||
return FALSE;
|
||||
}
|
||||
return TRUE;
|
||||
}
|
||||
|
||||
/** @brief
|
||||
Open database-handler.
|
||||
|
||||
|
|
@ -3681,6 +3653,18 @@ int handler::ha_write_row(uchar *buf)
|
|||
return 0;
|
||||
}
|
||||
|
||||
|
||||
/**
|
||||
Write a record to the engine bypassing row-level binary logging.
|
||||
This method is used internally by the server for writing to
|
||||
performance schema tables, which are never replicated.
|
||||
*/
|
||||
int handler::ha_write_row_no_binlog(uchar *buf)
|
||||
{
|
||||
return write_row(buf);
|
||||
}
|
||||
|
||||
|
||||
int handler::ha_update_row(const uchar *old_data, uchar *new_data)
|
||||
{
|
||||
int error;
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue