~mysqlatfacebook/mysqlatfacebook/5.0 : changes

~mysqlatfacebook/mysqlatfacebook/5.0 » Changes from revision 109

From Revision 109 to 90

Rev	Summary	Authors	Date
109	Update 5.0 to latest version of fast timer code from 5.1. Update 5.0 to latest version of fast timer code from 5.1. This version of the fast timer patch includes automatic detection of CPU synchronization. We need this in order to disable fast timers on hardware that does not prevent cycle counter drift between x86 cores.	Mark Callaghan	13 years ago
108	Initialize THD::examined_row_count Initialize THD::examined_row_count The slocket would show weird values for examined_row_count when showing operations that didn't examine rows, such as 'USE database'. This is because when no rows are examined, the variable is never set. This sets the variable when the THD is created, as is done in 5.1.	Mark Callaghan	13 years ago
107	Apply commit 97825 for bug 49238: Crash in row_drop_table_fo... Apply commit 97825 for bug 49238: Crash in row_drop_table_for_mysql(). Creating/Dropping a temporary table while at 1023 transactions will cause assert. Handle possible DB_TOO_MANY_CONCURRENT_TRXS when deleting metadata in row_drop_table_for_mysql().	Mark Callaghan	13 years ago
106	Re-enable fsync w/O_DIRECT when idb file changes size. Re-enable fsync w/O_DIRECT when idb file changes size. Some file systems don't write fs metadata immediately. After system crash the ibdata files appeared much smaller than they should have been preventing recovery. This change re-enables sync on O_DIRECT only when the file node size has changed since the last sync.	Mark Callaghan	13 years ago
105	Fix bug for my_fast_timer_is valid when low order 32 bits ar... Fix bug for my_fast_timer_is valid when low order 32 bits are 0	Mark Callaghan	13 years ago
104	Provide compile-time option to use gettimeofday instead of r... Provide compile-time option to use gettimeofday instead of rdtsc for my_fast_timer_t. Older x86 HW does not provide guarantees against clock drift for the cycle timer used by rdtsc. The only way to prevent this is to pin all threads to one core. A compile time option is added -- use ./configure C_EXTRA_FLAGS="-DMY_FAST_TIMER_RDTSC" to use rdtsc(). Otherwise, gettimeofday is used.	Mark Callaghan	13 years ago
103	Fix bug in slocket Lock_time and Lock_time fast calculations Fix bug in slocket Lock_time and Lock_time fast calculations The code used the difference between the wrong values when calculating lock times in the slocket. It used the time before the lock.	Mark Callaghan	13 years ago
102	Make innodb_lock_wait_timeout dynamic and per session. Make innodb_lock_wait_timeout dynamic and per session. Currently, innodb_lock_wait_timeout can only be set in my.cnf and not changed after mysqld is running. This changes that and allows it to be set per session.	Mark Callaghan	14 years ago
101	Add my.cnf option innodb_retry_io_on_error to retry read & w... Add my.cnf option innodb_retry_io_on_error to retry read & write once after a disk error (EIO). Add SHOW STATUS counters: Innodb_data_retried_reads - number of reads retried Innodb_data_retried_writes - number of writes retried Also print counters in SHOW INNODB STATUS in FILE section	Mark Callaghan	14 years ago
100	Add counters for innodb commit and rollback. Add to SHOW STA... Add counters for innodb commit and rollback. Add to SHOW STATUS Innodb_transaction_commit_all - counts all commits (readonly and readwrite) Innodb_transaction_commit_with_undo - count commits for which undo was created Innodb_transaction_rollback_total - count rollback of transaction Innodb_transaction_rollback_parital - count rollback to savepoint or from one statement Also, display output in TRANSACTION section of SHOW INNODB STATUS	Mark Callaghan	14 years ago
99	Add more attributes to slow query log. Add more attributes to slow query log. Query_time, Lock_time, Rows_sent and Rows_examined have always been there. All others are new and improve performance debugging. The output is: # Query_time: 0 Lock_time: 0 Rows_sent: 0 Rows_examined: 0 Thread_id: 3 Errno: 0 \ Killed: 0 Bytes_received: 110 Bytes_sent: 134 Read_first: 0 Read_last: 0 Read_key: 2 \ Read_next: 0 Read_prev: 0 Read_rnd: 0 Read_rnd_next: 0 Sort_merge_passes: 0 \ Sort_range_count: 0 Sort_rows: 0 Sort_scan_count: 0 Created_tmp_disk_tables: 0 \ Created_tmp_tables: 0 Start: 9:22:58 End: 9:22:58 Also, improve performance by moving some code out of the block in which the log mutex is locked. Note that I have not updated the equivalent code for the slow query log socket. Lets wait and see how this works before doing that.	Mark Callaghan	14 years ago
98	Fix deadlock from insert select on duplicate key update Fix deadlock from insert select on duplicate key update Fixes http://bugs.mysql.com/bug.php?id=51838	Mark Callaghan	14 years ago
97	Fix compiler warning in slocket code Fix compiler warning in slocket code	Mark Callaghan	14 years ago
96	wait_usecs_max was actually displaying svc_usecs_max wait_usecs_max was actually displaying svc_usecs_max	Mark Callaghan	14 years ago
95	Add a basic test of mysql slocket functionality: Add a basic test of mysql slocket functionality: - Verifies that slocket is disabled when there is no listener - Verifies that information is logged to slocket when there is a listener	Mark Callaghan	14 years ago
94	Port rpl_transaction_enabled from the Google patch. Port rpl_transaction_enabled from the Google patch. This makes slave replication state crash proof for changes to InnoDB and sometimes crash proof for other changes in the binlog. When this my.cnf parameter is set, the values in relay-log.info are also saved in the innodb transaction log and overwrite the values from the file during crash recovery. Messages are printed to the error log during startup: 100203 17:02:08 InnoDB: reading slave state from the transaction system header InnoDB: relay-log - filename /var/lib/mysql/mysqld-relay-bin.000159, position (19814802) InnoDB: master-log - filename binary-logs.000917, position (900814453) 100203 17:02:09 [Note] Read master.info: log_file_name: binary-logs.000917, pos: 1012638670 100203 17:02:09 [Note] init_relay_log_info: slave state from InnoDB: relay-log['/var/lib/mysql/mysqld-relay-bin.000159'(19814802)], master-log['binary -logs.000917' (900814453)] 100203 17:02:09 [Note] update_master_info: truncated file(/var/lib/mysql/mysqld-relay-bin.000159) from 131639038 to 131639019 100203 17:02:09 [Note] init_relay_log_info: slave state in relay-log.info: relay-log['/var/lib/mysql/mysqld-relay-bin.000159'(19814802)], master-log[' binary-logs.0 00917'(900814453)] 100203 17:02:09 [Note] Slave SQL thread initialized, starting replication in log 'binary-logs.000917' at position 900814453 and 900814453, relay log ' /var/lib/mysq l/mysqld-relay-bin.000159' position: 19814802 100203 17:02:09 [Note] Slave I/O thread: connected to master 'X', replication started in log 'binary-logs.000917' at position 1012638670 during shutdown: 100204 8:47:52 [Note] Slave I/O thread exiting, read up to log 'binary-logs.000923', position 150739714 100204 8:47:52 [Note] Error reading relay log event: slave SQL thread was killed 100204 8:47:52 [Note] Slave SQL thread set InnoDB state on shutdown 100204 8:47:52 [Note] InnoDB changed master log from 'binary-logs.000923' position 150739714 to 'binary-logs.000923' position 150739714 100204 8:47:52 [Note] InnoDB changed relay log from '/var/lib/mysql/mysqld-relay-bin.000178' position 150739898 to '/var/lib/mysql/mysqld-relay-bin.000178' position 150739898 The slave offset is written to InnoDB on shutdown. Prior to shutdown, transactions that don't use InnoDB (MyISAM for example) advance the slave offset without informing InnoDB. A mysqld crash will then recover slave state to the last commit done for InnoDB. This may then repeat the non InnoDB transactions.	Mark Callaghan	14 years ago
93	Create a unix datagram socket that when the log_datagram Create a unix datagram socket that when the log_datagram option is enabled, will recieve a log of all queries that take longer than log_datagram_usecs. The name for the socket is $datadir/slocket. The option can only be enabled when a client is listening on the socket.	Mark Callaghan	14 years ago
92	The SSL certs expired. Using patch from The SSL certs expired. Using patch from http://lists.mysql.com/commits/98627?f=plain to fix.	Mark Callaghan	14 years ago
91	Fix memory leak from innodb_thread_lifo patch Fix memory leak from innodb_thread_lifo patch	Mark Callaghan	14 years ago
90	Update to use fast timers for timing queries, and allow frac... Update to use fast timers for timing queries, and allow fractional (microsecond resolution) slow query time limit	Mark Callaghan	14 years ago

Older »