From 73e80314aa17b13d8bb0a6f0f230b99c439bee8a Mon Sep 17 00:00:00 2001 From: unknown Date: Sat, 24 May 2003 16:43:53 +0200 Subject: [PATCH 1/2] Fix for bug #490 and #491 (see details below) mysql-test/r/insert_select.result: Result update. mysql-test/r/rpl_insert_id.result: Test update mysql-test/t/insert_select.test: Check if a partly completed INSERT SELECT (failing because of "Duplicate key" after successfully inserting other rows) is written to the binlog if the table is not transactional and at least one row has been inserted (bug #491) mysql-test/t/rpl_insert_id.test: Test for bug #490 (INSERT SELECT in auto_increment) sql/sql_insert.cc: - In INSERT ... SELECT, if it fails with error but one row has been inserted and the table is not transactional, we must write to the binlog (the slave will stop because of the error code in the binlog event, this is normal). bug 491. - we must set INSERT_ID before writing to the binlog (bug 490 accidentally introduced by another dev in 4.0.13). --- mysql-test/r/insert_select.result | 14 +++++++++++++ mysql-test/r/rpl_insert_id.result | 28 ++++++++++++++++++++++++++ mysql-test/t/insert_select.test | 19 ++++++++++++++++++ mysql-test/t/rpl_insert_id.test | 33 +++++++++++++++++++++++++------ sql/sql_insert.cc | 24 +++++++++++++++++++--- 5 files changed, 109 insertions(+), 9 deletions(-) diff --git a/mysql-test/r/insert_select.result b/mysql-test/r/insert_select.result index e24c3179a0c..b49ca3a6c2f 100644 --- a/mysql-test/r/insert_select.result +++ b/mysql-test/r/insert_select.result @@ -66,3 +66,17 @@ INSERT INTO crash1 (numeropost,icone,contenu,pseudo,date,signature,ip) SELECT 1718,icone,contenu,pseudo,date,signature,ip FROM crash2 WHERE numeropost=9 ORDER BY numreponse ASC; DROP TABLE IF EXISTS crash1,crash2; +drop table if exists t1; +drop table if exists t2; +create table t1(a int, unique(a)); +insert into t1 values(2); +create table t2(a int); +insert into t2 values(1),(2); +reset master; +insert into t1 select * from t2; +Duplicate entry '2' for key 1 +show binlog events; +Log_name Pos Event_type Server_id Orig_log_pos Info +master-bin.001 4 Start 1 4 Server ver: VERSION, Binlog ver: 3 +master-bin.001 79 Query 1 79 use test; insert into t1 select * from t2 +drop table t1, t2; diff --git a/mysql-test/r/rpl_insert_id.result b/mysql-test/r/rpl_insert_id.result index d524818985e..6d996481a5c 100644 --- a/mysql-test/r/rpl_insert_id.result +++ b/mysql-test/r/rpl_insert_id.result @@ -39,3 +39,31 @@ b c 6 11 drop table t1; drop table t2; +create table t1(a int auto_increment, key(a)); +create table t2(b int auto_increment, c int, key(b)); +insert into t1 values (10); +insert into t1 values (null),(null),(null); +insert into t2 values (5,0); +insert into t2 (c) select * from t1; +select * from t2; +b c +5 0 +6 10 +7 11 +8 12 +9 13 +select * from t1; +a +10 +11 +12 +13 +select * from t2; +b c +5 0 +6 10 +7 11 +8 12 +9 13 +drop table t1; +drop table t2; diff --git a/mysql-test/t/insert_select.test b/mysql-test/t/insert_select.test index 42f65858d77..695f891f678 100644 --- a/mysql-test/t/insert_select.test +++ b/mysql-test/t/insert_select.test @@ -68,3 +68,22 @@ WHERE numeropost=9 ORDER BY numreponse ASC; DROP TABLE IF EXISTS crash1,crash2; + +# Addendum by Guilhem: +# Check if a partly-completed INSERT SELECT in a MyISAM table goes +# into the binlog +drop table if exists t1; +drop table if exists t2; +create table t1(a int, unique(a)); +insert into t1 values(2); +create table t2(a int); +insert into t2 values(1),(2); +reset master; +--error 1062 +insert into t1 select * from t2; +# The above should produce an error, but still be in the binlog; +# verify the binlog : +let $VERSION=`select version()`; +--replace_result $VERSION VERSION +show binlog events; +drop table t1, t2; diff --git a/mysql-test/t/rpl_insert_id.test b/mysql-test/t/rpl_insert_id.test index 3f3636d3082..3478aedf3eb 100644 --- a/mysql-test/t/rpl_insert_id.test +++ b/mysql-test/t/rpl_insert_id.test @@ -1,6 +1,7 @@ -#see if queries that use both -#auto_increment and LAST_INSERT_ID() -#are replicated well +# see if queries that use both +# auto_increment and LAST_INSERT_ID() +# are replicated well + source include/master-slave.inc; connection master; drop table if exists t1; @@ -15,9 +16,11 @@ sync_with_master; select * from t1; select * from t2; connection master; -#check if multi-line inserts, -#which set last_insert_id to the first id inserted, -#are replicated the same way + +# check if multi-line inserts, +# which set last_insert_id to the first id inserted, +# are replicated the same way + drop table t1; drop table t2; create table t1(a int auto_increment, key(a)); @@ -32,6 +35,24 @@ sync_with_master; select * from t1; select * from t2; connection master; + +# check if INSERT SELECT in auto_increment is well replicated (bug #490) + +drop table t1; +drop table t2; +create table t1(a int auto_increment, key(a)); +create table t2(b int auto_increment, c int, key(b)); +insert into t1 values (10); +insert into t1 values (null),(null),(null); +insert into t2 values (5,0); +insert into t2 (c) select * from t1; +select * from t2; +save_master_pos; +connection slave; +sync_with_master; +select * from t1; +select * from t2; +connection master; drop table t1; drop table t2; save_master_pos; diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc index 33a13fabdc6..167ccf974c7 100644 --- a/sql/sql_insert.cc +++ b/sql/sql_insert.cc @@ -1350,6 +1350,24 @@ void select_insert::send_error(uint errcode,const char *err) ::send_error(&thd->net,errcode,err); table->file->extra(HA_EXTRA_NO_CACHE); table->file->activate_all_index(thd); + /* + If at least one row has been inserted/modified and will stay in the table + (the table doesn't have transactions) (example: we got a duplicate key + error while inserting into a MyISAM table) we must write to the binlog (and + the error code will make the slave stop). + */ + if ((info.copied || info.deleted) && !table->file->has_transactions()) + { + if (last_insert_id) + thd->insert_id(last_insert_id); // For binary log + mysql_update_log.write(thd,thd->query,thd->query_length); + if (mysql_bin_log.is_open()) + { + Query_log_event qinfo(thd, thd->query, thd->query_length, + table->file->has_transactions()); + mysql_bin_log.write(&qinfo); + } + } ha_rollback_stmt(thd); if (info.copied || info.deleted) { @@ -1365,7 +1383,10 @@ bool select_insert::send_eof() error=table->file->activate_all_index(thd); table->file->extra(HA_EXTRA_NO_IGNORE_DUP_KEY); + if (last_insert_id) + thd->insert_id(last_insert_id); // For binary log /* Write to binlog before commiting transaction */ + mysql_update_log.write(thd,thd->query,thd->query_length); if (mysql_bin_log.is_open()) { Query_log_event qinfo(thd, thd->query, thd->query_length, @@ -1393,10 +1414,7 @@ bool select_insert::send_eof() else sprintf(buff,ER(ER_INSERT_INFO),info.records,info.deleted, thd->cuted_fields); - if (last_insert_id) - thd->insert_id(last_insert_id); // For update log ::send_ok(&thd->net,info.copied+info.deleted,last_insert_id,buff); - mysql_update_log.write(thd,thd->query,thd->query_length); return 0; } } From 2917fdb9d38c798cdc8e9341bec667ce4cb3c281 Mon Sep 17 00:00:00 2001 From: unknown Date: Sun, 25 May 2003 23:09:46 +0200 Subject: [PATCH 2/2] - Fix for memory leak if the SQL slave thread is killed just after reading an event. - A few more mutex locks and unlocks of rli.log_space_lock for doing clean reads of rli.ignore_log_space_limit - Broadcast after unlock, not before (small speed optimisation). --- sql/slave.cc | 29 +++++++++++++++++++++++++++-- 1 file changed, 27 insertions(+), 2 deletions(-) diff --git a/sql/slave.cc b/sql/slave.cc index e6215356ad1..b620603dc63 100644 --- a/sql/slave.cc +++ b/sql/slave.cc @@ -1322,6 +1322,8 @@ static bool wait_for_relay_log_space(RELAY_LOG_INFO* rli) !rli->ignore_log_space_limit) { pthread_cond_wait(&rli->log_space_cond, &rli->log_space_lock); + /* Re-acquire the mutex as pthread_cond_wait released it */ + pthread_mutex_lock(&rli->log_space_lock); } thd->proc_info = save_proc_info; pthread_mutex_unlock(&rli->log_space_lock); @@ -2028,7 +2030,11 @@ static int exec_relay_log_event(THD* thd, RELAY_LOG_INFO* rli) Log_event * ev = next_event(rli); DBUG_ASSERT(rli->sql_thd==thd); if (sql_slave_killed(thd,rli)) + { + /* do not forget to free ev ! */ + if (ev) delete ev; return 1; + } if (ev) { int type_code = ev->get_type_code(); @@ -2302,6 +2308,18 @@ reconnect done to recover from failed read"); goto err; } flush_master_info(mi); + /* + See if the relay logs take too much space. + We don't lock mi->rli.log_space_lock here; this dirty read saves time + and does not introduce any problem: + - if mi->rli.ignore_log_space_limit is 1 but becomes 0 just after (so + the clean value is 0), then we are reading only one more event as we + should, and we'll block only at the next event. No big deal. + - if mi->rli.ignore_log_space_limit is 0 but becomes 1 just after (so + the clean value is 1), then we are going into wait_for_relay_log_space() + for no reason, but this function will do a clean read, notice the clean + value and exit immediately. + */ if (mi->rli.log_space_limit && mi->rli.log_space_limit < mi->rli.log_space_total && !mi->rli.ignore_log_space_limit) @@ -2416,7 +2434,9 @@ slave_begin: rli->pending = 0; //tell the I/O thread to take relay_log_space_limit into account from now on + pthread_mutex_lock(&rli->log_space_lock); rli->ignore_log_space_limit= 0; + pthread_mutex_unlock(&rli->log_space_lock); if (init_relay_log_pos(rli, rli->relay_log_name, @@ -3122,9 +3142,14 @@ Log_event* next_event(RELAY_LOG_INFO* rli) pthread_mutex_lock(&rli->log_space_lock); // prevent the I/O thread from blocking next times rli->ignore_log_space_limit= 1; - // If the I/O thread is blocked, unblock it - pthread_cond_broadcast(&rli->log_space_cond); + /* + If the I/O thread is blocked, unblock it. + Ok to broadcast after unlock, because the mutex is only destroyed in + ~st_relay_log_info(), i.e. when rli is destroyed, and rli will not be + destroyed before we exit the present function. + */ pthread_mutex_unlock(&rli->log_space_lock); + pthread_cond_broadcast(&rli->log_space_cond); // Note that wait_for_update unlocks lock_log ! rli->relay_log.wait_for_update(rli->sql_thd); // re-acquire data lock since we released it earlier