From 2c8c15483d15d7a8e5b266ffcfbe7cf3503949f9 Mon Sep 17 00:00:00 2001 From: Monty Date: Fri, 16 Oct 2020 19:36:24 +0300 Subject: [PATCH] MDEV-23730 s3.replication_partition 'innodb,mix' segv This failure was caused because of several bugs: - Someone had removed s3-slave-ignore-updates=1 from slave.cnf, which caused the slave to remove files that the master was working on. - Bug in ha_partition::change_partitions() that didn't reset m_new_file in case of errors. This caused crashes in ha_maria::extra() as the maria handler was called on files that was already closed. - In ma_pagecache there was a bug that when one got a read error one a big block (s3 block), it left the flag PCBLOCK_BIG_READ on for the page which cased an assert when the page where flushed. - Flush all cached tables in case of ignored ALTER TABLE Note that when merging code from 10.3, that fixes the partition bug, use the code from this patch instead. Changes to ma_pagecache.cc written or reviewed by Sanja --- mysql-test/suite/parts/r/reorganize.result | 13 +++++++ mysql-test/suite/parts/t/reorganize.test | 20 +++++++++++ mysql-test/suite/s3/disabled.def | 3 -- .../suite/s3/replication_partition.result | 18 ++++++---- .../suite/s3/replication_partition.test | 22 +++++------- mysql-test/suite/s3/slave.cnf | 5 +++ sql/ha_partition.cc | 4 +++ sql/sql_table.cc | 35 +++++++++++++------ storage/maria/ma_pagecache.c | 11 +++--- 9 files changed, 94 insertions(+), 37 deletions(-) create mode 100644 mysql-test/suite/parts/r/reorganize.result create mode 100644 mysql-test/suite/parts/t/reorganize.test diff --git a/mysql-test/suite/parts/r/reorganize.result b/mysql-test/suite/parts/r/reorganize.result new file mode 100644 index 00000000000..5e6fe176edc --- /dev/null +++ b/mysql-test/suite/parts/r/reorganize.result @@ -0,0 +1,13 @@ +# +# MDEV-23248 Server crashes in mi_extra / +# ha_partition::loop_extra_alter upon REORGANIZE +# +CREATE TABLE t1 (a INT, b INT) ENGINE=MyISAM PARTITION BY RANGE (a) SUBPARTITION BY HASH (a) SUBPARTITIONS 70 (PARTITION p1 VALUES LESS THAN (100), PARTITION p2 VALUES LESS THAN MAXVALUE); +INSERT INTO t1 SELECT 4, 6 FROM seq_1_to_131072; +UPDATE t1 SET a = 7; +set @org_debug=@@debug_dbug; +set @@debug_dbug="+d,debug_abort_copy_partitions"; +ALTER TABLE t1 REORGANIZE PARTITION p1,p2 INTO (PARTITION p1 VALUES LESS THAN (5), PARTITION p2 VALUES LESS THAN MAXVALUE); +ERROR 42000: Table 't1' uses an extension that doesn't exist in this MariaDB version +set @@debug_dbug=@org_debug; +DROP TABLE t1; diff --git a/mysql-test/suite/parts/t/reorganize.test b/mysql-test/suite/parts/t/reorganize.test new file mode 100644 index 00000000000..66641254468 --- /dev/null +++ b/mysql-test/suite/parts/t/reorganize.test @@ -0,0 +1,20 @@ +--source include/have_sequence.inc +--source include/have_partition.inc +--source include/have_debug.inc + +--echo # +--echo # MDEV-23248 Server crashes in mi_extra / +--echo # ha_partition::loop_extra_alter upon REORGANIZE +--echo # + +CREATE TABLE t1 (a INT, b INT) ENGINE=MyISAM PARTITION BY RANGE (a) SUBPARTITION BY HASH (a) SUBPARTITIONS 70 (PARTITION p1 VALUES LESS THAN (100), PARTITION p2 VALUES LESS THAN MAXVALUE); +INSERT INTO t1 SELECT 4, 6 FROM seq_1_to_131072; +UPDATE t1 SET a = 7; + +set @org_debug=@@debug_dbug; +set @@debug_dbug="+d,debug_abort_copy_partitions"; +--error ER_UNSUPPORTED_EXTENSION +ALTER TABLE t1 REORGANIZE PARTITION p1,p2 INTO (PARTITION p1 VALUES LESS THAN (5), PARTITION p2 VALUES LESS THAN MAXVALUE); +set @@debug_dbug=@org_debug; + +DROP TABLE t1; diff --git a/mysql-test/suite/s3/disabled.def b/mysql-test/suite/s3/disabled.def index 8eae300e21e..e69de29bb2d 100644 --- a/mysql-test/suite/s3/disabled.def +++ b/mysql-test/suite/s3/disabled.def @@ -1,3 +0,0 @@ -replication_partition : MDEV-23730: Server crashes in ha_maria::extra -replication_mixed : MDEV-23770: Replication failure -replication_stmt : MDEV-23770: Replication failure diff --git a/mysql-test/suite/s3/replication_partition.result b/mysql-test/suite/s3/replication_partition.result index 40fb2aec698..2b9297ea9a0 100644 --- a/mysql-test/suite/s3/replication_partition.result +++ b/mysql-test/suite/s3/replication_partition.result @@ -89,10 +89,18 @@ select sum(c1)+0 from t1; sum(c1)+0 600 stop slave; -flush tables; -select sum(c1)+0 from t1; -sum(c1)+0 -600 +show create table t1; +Table Create Table +t1 CREATE TABLE `t1` ( + `c1` int(11) NOT NULL, + `c2` int(11) DEFAULT NULL, + PRIMARY KEY (`c1`) +) ENGINE=S3 DEFAULT CHARSET=latin1 + PARTITION BY RANGE (`c1`) +(PARTITION `p1` VALUES LESS THAN (200) ENGINE = S3, + PARTITION `p2` VALUES LESS THAN (300) ENGINE = S3, + PARTITION `p3` VALUES LESS THAN (400) ENGINE = S3, + PARTITION `p4` VALUES LESS THAN (500) ENGINE = S3) connection master; drop table t1; connection slave; @@ -204,8 +212,6 @@ slave-bin.000001 # Query # # use `database`; alter table t1 engine=S3 slave-bin.000001 # Gtid # # GTID #-#-# slave-bin.000001 # Query # # use `database`; set @@sql_if_exists=1; ALTER TABLE t1 ADD PARTITION (PARTITION p4 VALUES LESS THAN (500)) slave-bin.000001 # Gtid # # GTID #-#-# -slave-bin.000001 # Query # # use `database`; flush tables -slave-bin.000001 # Gtid # # GTID #-#-# slave-bin.000001 # Query # # use `database`; DROP TABLE IF EXISTS `t1` /* generated by server */ slave-bin.000001 # Gtid # # GTID #-#-# slave-bin.000001 # Query # # use `database`; CREATE TABLE t2 ( diff --git a/mysql-test/suite/s3/replication_partition.test b/mysql-test/suite/s3/replication_partition.test index 0430d14d7a7..8a8699920bf 100644 --- a/mysql-test/suite/s3/replication_partition.test +++ b/mysql-test/suite/s3/replication_partition.test @@ -7,6 +7,12 @@ --source create_database.inc sync_slave_with_master; + +if (`select @@s3_slave_ignore_updates <> 1`) +{ + die "Slave is not configured with s3-slave-ignore-updates=1"; +} + let $MYSQLD_DATADIR= `select @@datadir`; --replace_result $database database --eval use $database @@ -65,20 +71,8 @@ sync_slave_with_master; select sum(c1)+0 from t1; stop slave; -# .frm amd .par files should not exists on the salve as it has just seen the -# ALTER TABLE which cased the removal of the .frm and .par files. The table -# from the above "select sum()" came from table cache and was used as it's -# id matches the one in S3 ---error 1 ---file_exists $MYSQLD_DATADIR/$database/t1.frm ---error 1 ---file_exists $MYSQLD_DATADIR/$database/t1.par -# Flushing the table cache will force the .frm and .par files to be -# re-generated -flush tables; -select sum(c1)+0 from t1; ---file_exists $MYSQLD_DATADIR/$database/t1.frm ---file_exists $MYSQLD_DATADIR/$database/t1.par +# Ensure the slave is using the new table +show create table t1; connection master; drop table t1; diff --git a/mysql-test/suite/s3/slave.cnf b/mysql-test/suite/s3/slave.cnf index b9b1da73a92..ade80c9ef8e 100644 --- a/mysql-test/suite/s3/slave.cnf +++ b/mysql-test/suite/s3/slave.cnf @@ -2,6 +2,11 @@ plugin-maturity = alpha plugin-load-add=@ENV.HA_S3_SO s3=ON +s3-slave-ignore-updates=1 + +# You can change the following when running the tests against +# your own S3 setup + #s3-host-name=s3.amazonaws.com #s3-protocol-version=Amazon #s3-bucket=MariaDB diff --git a/sql/ha_partition.cc b/sql/ha_partition.cc index 921b461f15e..a3c0029adbb 100644 --- a/sql/ha_partition.cc +++ b/sql/ha_partition.cc @@ -2069,6 +2069,7 @@ int ha_partition::change_partitions(HA_CREATE_INFO *create_info, DBUG_ASSERT(part_elem->part_state == PART_TO_BE_REORGED); part_elem->part_state= PART_TO_BE_DROPPED; } + DBUG_ASSERT(m_new_file == 0); m_new_file= new_file_array; if (unlikely((error= copy_partitions(copied, deleted)))) { @@ -2077,6 +2078,7 @@ int ha_partition::change_partitions(HA_CREATE_INFO *create_info, They will later be deleted through the ddl-log. */ cleanup_new_partition(part_count); + m_new_file= 0; } DBUG_RETURN(error); } @@ -2166,6 +2168,8 @@ int ha_partition::copy_partitions(ulonglong * const copied, file->ha_rnd_end(); reorg_part++; } + DBUG_EXECUTE_IF("debug_abort_copy_partitions", + DBUG_RETURN(HA_ERR_UNSUPPORTED); ); DBUG_RETURN(FALSE); error: m_reorged_file[reorg_part]->ha_rnd_end(); diff --git a/sql/sql_table.cc b/sql/sql_table.cc index a1c58b53bb0..8db2ec3d761 100644 --- a/sql/sql_table.cc +++ b/sql/sql_table.cc @@ -9739,6 +9739,22 @@ static int create_table_for_inplace_alter(THD *thd, } +/* + log query if slave thread and send my_ok() + + Help function for mysql_alter_table() +*/ + +static bool log_and_ok(THD *thd) +{ + if (thd->slave_thread && + write_bin_log(thd, true, thd->query(), thd->query_length())) + return(true); + my_ok(thd); + return(0); +} + + /** Alter table @@ -9873,10 +9889,7 @@ bool mysql_alter_table(THD *thd, const LEX_CSTRING *new_db, table_list->mdl_request.type= MDL_EXCLUSIVE; /* This will only drop the .frm file and local tables, not shared ones */ error= mysql_rm_table(thd, table_list, 1, 0, 0, 1); - if (write_bin_log(thd, true, thd->query(), thd->query_length()) || error) - DBUG_RETURN(true); - my_ok(thd); - DBUG_RETURN(0); + DBUG_RETURN(log_and_ok(thd)); } /* @@ -9908,16 +9921,14 @@ bool mysql_alter_table(THD *thd, const LEX_CSTRING *new_db, the statement as this slave may not have the table shared */ thd->clear_error(); - if (thd->slave_thread && - write_bin_log(thd, true, thd->query(), thd->query_length())) - DBUG_RETURN(true); - my_ok(thd); - DBUG_RETURN(0); + DBUG_RETURN(log_and_ok(thd)); } } DBUG_RETURN(true); } + table= table_list->table; + #ifdef WITH_WSREP if (WSREP(thd) && (thd->lex->sql_command == SQLCOM_ALTER_TABLE || @@ -9929,7 +9940,6 @@ bool mysql_alter_table(THD *thd, const LEX_CSTRING *new_db, DEBUG_SYNC(thd, "alter_table_after_open_tables"); - table= table_list->table; if (table->versioned()) { if (handlerton *hton1= create_info->db_type) @@ -9968,12 +9978,17 @@ bool mysql_alter_table(THD *thd, const LEX_CSTRING *new_db, Alter_table_ctx alter_ctx(thd, table_list, tables_opened, new_db, new_name); mdl_ticket= table->mdl_ticket; + /* + We have to do a check also after table is opened as there could be no + ENGINE= on the command line or the table could a partitioned S3 table. + */ if (table->file->check_if_updates_are_ignored("ALTER")) { /* Table is a shared table. Remove the .frm file. Discovery will create a new one if needed. */ + table->s->tdc->flushed= 1; // Force close of all instances if (thd->mdl_context.upgrade_shared_lock(mdl_ticket, MDL_EXCLUSIVE, thd->variables.lock_wait_timeout)) diff --git a/storage/maria/ma_pagecache.c b/storage/maria/ma_pagecache.c index 0a55018db0f..bf646115bd9 100644 --- a/storage/maria/ma_pagecache.c +++ b/storage/maria/ma_pagecache.c @@ -2810,7 +2810,7 @@ static void read_big_block(PAGECACHE *pagecache, if (block_to_read->status & PCBLOCK_ERROR) { /* We get first block with an error so all operation failed */ - goto error; + DBUG_VOID_RETURN; } if (block_to_read->status & PCBLOCK_BIG_READ) { @@ -2842,7 +2842,11 @@ static void read_big_block(PAGECACHE *pagecache, // page should be read by other thread DBUG_ASSERT(block->status & PCBLOCK_READ || block->status & PCBLOCK_ERROR); - DBUG_ASSERT(block->status & PCBLOCK_BIG_READ); + /* + It is possible that other thread already removed the flag (in + case of two threads waiting) but it will not make harm to try to + remove it even in that case. + */ block->status&= ~PCBLOCK_BIG_READ; // all is read => lets finish nice DBUG_ASSERT(block_to_read != block); @@ -2960,9 +2964,8 @@ static void read_big_block(PAGECACHE *pagecache, } pagecache->big_block_free(&data); - block_to_read->status&= ~PCBLOCK_BIG_READ; - end: + block_to_read->status&= ~PCBLOCK_BIG_READ; if (block_to_read != block) { remove_reader(block_to_read);