database/mariadb
1
0
Fork 0
mirror of https://github.com/MariaDB/server.git synced 2025-07-29 05:21:33 +03:00
Code Activity

Merge 10.1 into 10.2

Browse Source
This commit is contained in:
Marko Mäkelä
2018-12-28 17:40:38 +02:00
parent 734029fa79 c5a5eaa9a9
commit 33caaba5c8
5 changed files with 82 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
Docs/README-wsrep
View File

@ -60,7 +60,7 @@ CONTENTS:
Wsrep API developed by Codership Oy is a modern generic (database-agnostic)
replication API for transactional databases with a goal to make database
replication/logging subsystem completely modular and pluggable. It is developed
with flexibility and completeness in mind to satisfy broad range of modern
with flexibility and completeness in mind to satisfy a broad range of modern
replication scenarios. It is equally suitable for synchronous and asynchronous,
master-slave and multi-master replication.
@ -87,7 +87,7 @@ Upgrade from mysql-server-5.0 to mysql-wsrep is not supported yet, please
upgrade to mysql-server-5.1 first.
If you're installing over an existing mysql installation, mysql-server-wsrep
will conflict with mysql-server-5.1 package, so remove it first:
will conflict with the mysql-server-5.1 package, so remove it first:
$ sudo apt-get remove mysql-server-5.1 mysql-server-core-5.1
@ -105,7 +105,7 @@ For example, installation of required packages on Debian Lenny:
$ sudo apt-get install psmisc
$ sudo apt-get -t lenny-backports install mysql-client-5.1
Now you should be able to install mysql-wsrep package:
Now you should be able to install the mysql-wsrep package:
$ sudo dpkg -i <mysql-server-wsrep DEB>
@ -150,7 +150,7 @@ and can be ignored unless specific functionality is needed.
3. FIRST TIME SETUP
Unless you're upgrading an already installed mysql-wsrep package, you will need
to set up a few things to prepare server for operation.
to set up a few things to prepare the server for operation.
3.1 CONFIGURATION FILES
@ -162,7 +162,7 @@ to set up a few things to prepare server for operation.
* Make sure system-wide my.cnf contains "!includedir /etc/mysql/conf.d/" line.
* Edit /etc/mysql/conf.d/wsrep.cnf and set wsrep_provider option by specifying
a path to provider library. If you don't have a provider, leave it as it is.
a path to the provider library. If you don't have a provider, leave it as it is.
* When a new node joins the cluster it'll have to receive a state snapshot from
one of the peers. This requires a privileged MySQL account with access from
@ -267,7 +267,7 @@ innodb_autoinc_lock_mode=2
This is a required parameter. Without it INSERTs into tables with
AUTO_INCREMENT column may fail.
autoinc lock modes 0 and 1 can cause unresolved deadlock, and make
system unresponsive.
the system unresponsive.
innodb_locks_unsafe_for_binlog=1
This option is required for parallel applying.
@ -299,14 +299,14 @@ wsrep_node_address=
results (multiple network interfaces, NAT, etc.)
If not explicitly overridden by wsrep_sst_receive_address, the <address> part
will be used to listen for SST (see below). And the whole <address>[:port]
will be passed to wsrep provider to be used as a base address in its
will be passed to the wsrep provider to be used as a base address in its
communications.
wsrep_node_name=
Human readable node name (for easier log reading only). Defaults to hostname.
wsrep_slave_threads=1
Number of threads dedicated to processing of writesets from other nodes.
The number of threads dedicated to the processing of writesets from other nodes.
For best performance should be few per CPU core.
wsrep_dbug_option
@ -326,7 +326,7 @@ wsrep_convert_LOCK_to_trx=0
wsrep_retry_autocommit=1
Retry autocommit queries and single statement transactions should they fail
certification test. This is analogous to rescheduling an autocommit query
should it go into deadlock with other transactions in the database lock
should it go into a deadlock with other transactions in the database lock
manager.
wsrep_auto_increment_control=1
@ -357,7 +357,7 @@ wsrep_OSU_method=TOI
is not replicating and may be unable to process replication events (due to
table lock). Once DDL operation is complete, the node will catch up and sync
with the cluster to become fully operational again. The DDL statement or
its effects are not replicated, so it is user's responsibility to manually
its effects are not replicated, so it is the user's responsibility to manually
perform this operation on each of the nodes.
wsrep_forced_binlog_format=none
@ -366,7 +366,7 @@ wsrep_forced_binlog_format=none
format, regardless of what the client session has specified in binlog_format.
Valid choices for wsrep_forced_binlog_format are: ROW, STATEMENT, MIXED and
special value NONE, meaning that there is no forced binlog format in effect.
This variable was intruduced to support STATEMENT format replication during
This variable was introduced to support STATEMENT format replication during
rolling schema upgrade processing. However, in most cases ROW replication
is valid for asymmetrict schema replication.
@ -412,8 +412,8 @@ wsrep_sst_auth=
wsrep_sst_donor=
A name of the node which should serve as state snapshot donor. This allows
to control which node will serve state snapshot request. By default the
most suitable node is chosen by wsrep provider. This is the same as given in
controlling which node will serve the state snapshot request. By default the
most suitable node is chosen by the wsrep provider. This is the same as given in
wsrep_node_name.
@ -423,7 +423,7 @@ wsrep_sst_donor=
for the database. They change the database structure and are non-
transactional.
Release 22.3 brings a new method for performing schema upgrades. User can
Release 22.3 brings a new method for performing schema upgrades. A user can
now choose whether to use the traditional total order isolation or new
rolling schema upgrade method. The OSU method choice is done by global
parameter: 'wsrep_OSU_method'.
@ -439,7 +439,7 @@ wsrep_sst_donor=
6.2 Rolling Schema Upgrade (RSU)
Rolling schema upgrade is new DDL processing method, where DDL will be
Rolling schema upgrade is a new DDL processing method, where DDL will be
processed locally for the node. The node is disconnected of the replication
for the duration of the DDL processing, so that there is only DDL statement
processing in the node and it does not block the rest of the cluster. When
@ -468,7 +468,7 @@ wsrep_sst_donor=
* LOCK/UNLOCK TABLES cannot be supported in multi-master setups.
* lock functions (GET_LOCK(), RELEASE_LOCK()... )
4) Query log cannot be directed to table. If you enable query logging,
4) Query log cannot be directed to a table. If you enable query logging,
you must forward the log to a file:
log_output = FILE
Use general_log and general_log_file to choose query logging and the
@ -480,7 +480,7 @@ wsrep_sst_donor=
6) Due to cluster level optimistic concurrency control, transaction issuing
COMMIT may still be aborted at that stage. There can be two transactions.
writing to same rows and committing in separate cluster nodes, and only one
of the them can successfully commit. The failing one will be aborted.
of them can successfully commit. The failing one will be aborted.
For cluster level aborts, MySQL/galera cluster gives back deadlock error.
code (Error: 1213 SQLSTATE: 40001 (ER_LOCK_DEADLOCK)).

24
mysql-test/suite/innodb/r/innodb-alter-debug.result
View File

@ -75,3 +75,27 @@ connection default;
ERROR 23000: Duplicate entry '1' for key 'a'
SET DEBUG_SYNC='RESET';
DROP TABLE t1;
#
# MDEV-17470 Orphan temporary files after interrupted ALTER
# cause InnoDB: Operating system error number 17 and eventual
# fatal error 71
#
CREATE TABLE t1 (pk INT AUTO_INCREMENT PRIMARY KEY, i INT) ENGINE=InnoDB;
INSERT INTO t1 VALUES (NULL,1),(NULL,2),(NULL,3),(NULL,4),(NULL,5),(NULL,6),(NULL,7),(NULL,8);
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
LOCK TABLE t1 READ;
connect con1,localhost,root,,test;
SET max_statement_time= 1;
ALTER TABLE t1 FORCE, ALGORITHM=COPY;
ERROR 70100: Query execution was interrupted (max_statement_time exceeded)
SET DEBUG_SYNC = 'now SIGNAL stop_waining';
disconnect con1;
connection default;
SET DEBUG_SYNC = 'now WAIT_FOR stop_waining';
UNLOCK TABLES;
DROP TABLE t1;
SET DEBUG_SYNC = 'RESET';

28
mysql-test/suite/innodb/t/innodb-alter-debug.test
View File

@ -100,3 +100,31 @@ DROP TABLE t1;
# Wait till all disconnects are completed
--source include/wait_until_count_sessions.inc
--echo #
--echo # MDEV-17470 Orphan temporary files after interrupted ALTER
--echo # cause InnoDB: Operating system error number 17 and eventual
--echo # fatal error 71
--echo #
CREATE TABLE t1 (pk INT AUTO_INCREMENT PRIMARY KEY, i INT) ENGINE=InnoDB;
INSERT INTO t1 VALUES (NULL,1),(NULL,2),(NULL,3),(NULL,4),(NULL,5),(NULL,6),(NULL,7),(NULL,8);
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
INSERT INTO t1 SELECT NULL, i FROM t1;
LOCK TABLE t1 READ;
--connect (con1,localhost,root,,test)
SET max_statement_time= 1;
--error ER_STATEMENT_TIMEOUT
ALTER TABLE t1 FORCE, ALGORITHM=COPY;
SET DEBUG_SYNC = 'now SIGNAL stop_waining';
--disconnect con1
--connection default
SET DEBUG_SYNC = 'now WAIT_FOR stop_waining';
UNLOCK TABLES;
DROP TABLE t1;
SET DEBUG_SYNC = 'RESET';

16
storage/innobase/row/row0mysql.cc
View File

@ -3273,7 +3273,6 @@ row_drop_table_for_mysql(
char* tablename = NULL;
bool locked_dictionary = false;
pars_info_t* info = NULL;
mem_heap_t* heap = NULL;
DBUG_ENTER("row_drop_table_for_mysql");
DBUG_PRINT("row_drop_table_for_mysql", ("table: '%s'", name));
@ -3305,8 +3304,11 @@ row_drop_table_for_mysql(
| DICT_ERR_IGNORE_CORRUPT));
if (!table) {
err = DB_TABLE_NOT_FOUND;
goto funct_exit;
if (locked_dictionary) {
row_mysql_unlock_data_dictionary(trx);
}
trx->op_info = "";
DBUG_RETURN(DB_TABLE_NOT_FOUND);
}
/* This function is called recursively via fts_drop_tables(). */
@ -3357,6 +3359,9 @@ row_drop_table_for_mysql(
/* make sure background stats thread is not running on the table */
ut_ad(!(table->stats_bg_flag & BG_STAT_IN_PROGRESS));
const bool is_temp_name = strstr(table->name.m_name,
"/" TEMP_FILE_PREFIX);
mem_heap_t* heap = NULL;
if (!dict_table_is_temporary(table)) {
if (table->space != TRX_SYS_SPACE) {
@ -3373,7 +3378,7 @@ row_drop_table_for_mysql(
hold the InnoDB dictionary lock, we will drop any
adaptive hash index entries upfront. */
while (buf_LRU_drop_page_hash_for_tablespace(table)) {
if (trx_is_interrupted(trx)
if ((!is_temp_name && trx_is_interrupted(trx))
|| srv_shutdown_state
!= SRV_SHUTDOWN_NONE) {
err = DB_INTERRUPTED;
@ -3464,7 +3469,6 @@ row_drop_table_for_mysql(
}
}
DBUG_EXECUTE_IF("row_drop_table_add_to_background", goto defer;);
/* TODO: could we replace the counter n_foreign_key_checks_running
@ -3475,7 +3479,7 @@ row_drop_table_for_mysql(
if (table->n_foreign_key_checks_running > 0) {
defer:
if (!strstr(table->name.m_name, "/" TEMP_FILE_PREFIX)) {
if (!is_temp_name) {
heap = mem_heap_create(FN_REFLEN);
const char* tmp_name
= dict_mem_create_temporary_tablename(

5
storage/xtradb/row/row0mysql.cc
View File

@ -4234,9 +4234,10 @@ row_drop_table_for_mysql(
calling btr_search_drop_page_hash_index() while we
hold the InnoDB dictionary lock, we will drop any
adaptive hash index entries upfront. */
const bool is_temp = dict_table_is_temporary(table)
|| strstr(tablename_minus_db, tmp_file_prefix);
while (buf_LRU_drop_page_hash_for_tablespace(table)) {
if ((!dict_table_is_temporary(table)
&& trx_is_interrupted(trx))
if ((!is_temp && trx_is_interrupted(trx))
|| srv_shutdown_state != SRV_SHUTDOWN_NONE) {
err = DB_INTERRUPTED;
goto funct_exit;