MySQL Advanced
MySQL Replication
MySQL Cluster
MySQL Partitioning
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Advanced
MySQL Replication
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Replication
Concepts
MySQL replication tracks updates performed in a MySQL master
logs and updates one or more slave servers. The mechanism
involved in MySQL replication requires binary logging to be
active.
-Asynchronous replication, a slave doesn't need to be always
active.
-MySQL replication can be used to scale-out queries among
different servers.
-MySQL replication is a good solution for performing live
backups.
- Long distance data distribution.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Replication
Architecture
MySQL replication architecture:
- Each update performed in a database is replicated from the master to
the slaves.
- Only slaves keep track of their replication status.
- One may set at what point the replication begins.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Replication
setup
MySQL Master
MySQL slave
Turn on binary log and set the
server id restart mysql server:
Set the master properties in
/etc/mysql/my.cnf:
server-id =1
server-id=2

log_bin =/var/log/mysql/mysql-bin.log master-host = master_ip
binlog_do_db = gaia_test
Create a user for replication:

GRANT REPLICATION SLAVE ON *.*
TO 'rep_slave'@'slave_ip' IDENTIFIED
BY 'password';
master-user = rep_slave
master-password = password
replicate-do-db={db_name}
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Replication
setup
MySQL Master
MySQL slave
•Show master log position:
•Change log position:
mysql > show master status \G;
CHANGE MASTER TO
File: mysql-bin.000011
Position: 784
Binlog_Do_DB: gaia_test
MASTER_HOST=server_IP,
MASTER_LOG_FILE=mysqlbin,
MASTER_LOG_POS=784;
Start Slave:
mysql > slave start;
•See slave status and notice
the SLAVE_* variables for
errors (No = bad) :
Show slave status\G;
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Advanced
MySQL Cluster
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
MySQL cluster is designed to provide near perfect availability
through a shared nothing architecture, in turn this means that all
the nodes in a MySQL cluster can be replicated for failover.
MySQL cluster is a in memory database, which means that all
data must be located in RAM, since version 5.1.6 a mechanism is
available to store data on disk although Indexes must still fit in
ram.
All data in MySQL is horizontally partitioned, across different
storage nodes, which means that all table data is split across the
storage nodes.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
MySQL cluster architecture consists of three types nodes:
Storage nodes:
All data is stored in this nodes.
SQL nodes:
Standard MySQL nodes, used for connection.
Management nodes:
Used to change the setup of the cluster.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
Setup (Debian)
To setup a 3 node (API + engine + management) MySQL
Cluster you must update 2 files:
Edit /etc/mysql/my.cnf and add:
[mysqld]
ndbcluster
[MYSQL_CLUSTER]
ndb-connectstring=127.0.0.1
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
Setup (debian)
Create a new configuration file in /etc/mysql/ndb_mgmd.cnf:
[ndbd default]
NoOfReplicas=1
DataMemory=80M
IndexMemory=18M
[ndb_mgmd]
Id=1
hostname=localhost
datadir=/var/lib/mysql-cluster
[mysqld]
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
Startup order:
#1 Management nodes:
invoke-rc.d mysql-ndb-mgm start (start-initial)
#2 Data nodes
invoke-rc.d mysql-ndb start
#3 SQL nodes
invoke-rc.d mysql start
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
Testing the installation:
bash$ ndb-mgm -e show
Connected to Management Server at: 127.0.0.1:1186
Cluster Configuration
---------------------
[ndbd(NDB)]
id=2
1 node(s)
@127.0.0.1 (Version: 5.1.22, Nodegroup: 0, Master)
[ndb_mgmd(MGM)] 1 node(s)
id=1
@127.0.0.1 (Version: 5.1.22)
[mysqld(API)]
id=3
1 node(s)
@127.0.0.1 (Version: 5.1.22)
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Cluster
Testing the installation:
mysql> create table person (id int, name varchar(32)) engine = ndb;
Query OK, 0 rows affected (2,34 sec)
mysql> insert into person (id, name ) values ('11','carlos');
Query OK, 1 row affected (0,00 sec)
mysql> select * from person;
+------+--------+
| id
| name
|
+------+--------+
|
11 | carlos |
+------+--------+
1 row in set (0,00 sec)
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Advanced
MySQL Partitioning
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
concepts
Typical SQL doesn't provide guidance to underling physical
storage layer, Partitioning allows you to split a table across
multiple files, through the usage of a special partitioning rule
called partitioning function.
Partitioning can only take place in engines that support
partitioning, engines like MyISAM or InnoDB and not like CSV or
BlackHole.
The big performance boost in database partition is that partitions
that do not satisfy a certain rule are not scanned, this is called
partition pruning.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
Partition types
• RANGE partitioning: Assigns rows to partitions based on column
values falling within a given range.
• LIST partitioning: Similar to partitioning by range, except that
the partition is selected based on columns matching one of a
Partitioning set of discrete values.
• HASH partitioning: A partition is selected based on the value
returned by a user-defined expression that operates on column
values in rows to be inserted into the table. The function may
consist of any expression valid in MySQL that yields a nonnegative integer value.
• KEY partitioning: Similar to partitioning by hash, except that
only one or more columns to be evaluated are supplied, and the
MySQL server provides its own hashing function.
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
examples
Range partition:
CREATE TABLE t1 (
r_name VARCHAR(50) NOT NULL,
region_code TINYINT UNSIGNED NOT NULL
)
PARTITION BY RANGE( region_code ) (
PARTITION p0 VALUES LESS THAN (64),
PARTITION p1 VALUES LESS THAN (128),
PARTITION p3 VALUES LESS THAN MAXVALUE
);
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
examples
List Partition:
CREATE TABLE employees (
id INT NOT NULL,
fname VARCHAR(30),
lname VARCHAR(30),
hired DATE NOT NULL DEFAULT '1970-01-01',
separated DATE NOT NULL DEFAULT '9999-12-31',
job_code INT,
store_id INT )
PARTITION BY LIST(store_id) (
PARTITION pNorth VALUES IN (16,15,3,5,6,9,17),
PARTITION pEast VALUES IN (1,2,8,10,11,19,20),
PARTITION pWest VALUES IN (7, 4,12,13,14,18) );
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
examples
Hash Partition:
CREATE TABLE employees (
id INT NOT NULL,
fname VARCHAR(30),
lname VARCHAR(30),
hired DATE NOT NULL DEFAULT '1970-01-01',
separated DATE NOT NULL DEFAULT '9999-12-31',
job_code INT,
store_id INT )
PARTITION BY HASH(store_id)
PARTITIONS 5;
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
MySQL Partitioning
examples
Key Partition:
CREATE TABLE k1 (
id INT NOT NULL PRIMARY KEY,
name VARCHAR(20))
PARTITION BY KEY()
PARTITIONS 4;
CREATE TABLE k1 (
id INT NOT NULL,
name VARCHAR(20),
UNIQUE KEY (id) )
PARTITION BY KEY()
PARTITIONS 4;
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
Q&A
•For more information:
More on MySQL-cluster and other studies:
SIM Studies at the GAIA WIKI
The technical note:
GAIA-C1-TN-SIM-CDJ-001-1.pdf
GAIA-C1-TN-SIM-AAB-001-01.pdf
António Amorim, Carlos Jesus. CU1 - First Database Testing Meeting, 20/Nov/08
MySQL Benchmarks
MySQL Single server
MySQL Cluster
Oracle
António Amorim, Carlos Jesus. CU1 - DBWorkshop 9-10 /Feb/09
Test objectives
•
Past studies evaluated the architecture and performance of MySQL Cluster and PostgreSQL
solutions for the Gaia Databases.
•
MySQL Cluster:
–
MySQL-Cluster
•
•
–
Poor performance
Redundancy benefits
PostgreSQL
•
•
Sends 2x more data than MySQL
Poor Bytea performance
Network load(400K objects)
PostgreSQL: 2072M
MYSQL:
1050 Mb
ORACLE-XE: 520 Mb
António Amorim, Carlos Jesus. CU1 - First Database Testing Meeting, 20/Nov/08
Single MySQL Server
In order to pinpoint bottlenecks associated with MySQL, a series of basic tests were performed, these new
tests made use of the available MDBExtractorIngestor package.
Control and Monitor:
Available at the local infrastructure was the SMS (Supervisor Monitor Scheduler) batch system, we deployed in one node the database and in another the
Ingestor.
At the local infrastructure was available the SMS (Supervisor Monitor Scheduler) batch system.
The tests basic 2 node client server model the nodes where as following:
Hardware setup
António Amorim, Carlos Jesus. CU1 - First Database Testing Meeting, 20/Nov/08
Tests For
MDBExtractorIngestor
JdbcObjectUpdater.java
MySQL Ingestor, Old and new JdbcObjectUpdater, 4.5Gb of data.
António Amorim, Carlos Jesus. CU1 - First Database Testing Meeting, 20/Nov/08
Tests For
DbBenchmark
JdbcObjectUpdater.java
MySQL Ingestor, Old and new JdbcObjectUpdater, 1Gb of data.
Oracle datapump (local), 1Gb of data.
António Amorim, Carlos Jesus. CU1 - First Database Testing Meeting, 20/Nov/08