The Database Solution: Load balancing PXC with ProxySQL

ProxySQL is widely used load balancing tool for HA MySQL load balancing. When we have HA clusters it is always better to use a load balancer for scaling the database. ie, which nodes can be used for writes and who for reads.

In this article, I will explain the below

⇒ Installing , Starting ProxySQL

⇒ How Percona XtraDB Cluster (PXC) can be used with ProxySQL with READ-WRITE split.

⇒ Read/write benchmark test

⇒ Automatic failover.

To read more about ProxySQL click here. {https://github.com/sysown/proxysql/wiki}

For this setup i have the below,

Linux Server : Centos 6.10

PXC Version : 5.7.25-28-57 Percona XtraDB Cluster (GPL),

ProxySQL version : ProxySQL 1.4.4

Proxy Node : 192.168.56.106

Galera Node1 : 192.168.56.102

Galera Node1 : 192.168.56.103

Galera Node1 : 192.168.56.107

Below is my architecture,

PXC setup can be done from this doc,

Installing ProxySQL

The best way to install proxysql is from Percona Repo. To achieve this, install the percona repo on the server initially (if not available)

Note : The following commands are run under root user , if not root user please use sudo before the commands

yum install https://repo.percona.com/yum/percona-release-latest.noarch.rpm

You can get the rpm from below command,

[root@centos ~]# yum list | grep proxysql

proxysql.x86_64 1.4.14-1.1.el6 @percona-release-x86_64

[root@centos ~]#

To install proxysql run ,

yum install proxysql

You can even check the same post installation

[root@centos ~]# rpm -qa | grep proxysql

proxysql-1.4.14-1.1.el6.x86_64

[root@centos ~]#

Note : You need to have mysql-client to be installed on the same proxy node.

Starting ProxySQL

You need to start the service ,

service proxysql start

/etc/init.d/proxysql start

ProxySQL will create two config files where your configurations are available

proxysql-admin.cnf & proxysql.cnf

In proxysql-admin.cnf you can have all your admin variables , below is what i had

cat /etc/proxysql-admin.cnf

# proxysql admin interface credentials.

export PROXYSQL_DATADIR='/var/lib/proxysql'

export PROXYSQL_USERNAME='admin'

export PROXYSQL_PASSWORD='admin'

export PROXYSQL_HOSTNAME='localhost'

export PROXYSQL_PORT='6032'

# PXC admin credentials for connecting to pxc-cluster-node.

export CLUSTER_USERNAME='admin'

export CLUSTER_PASSWORD='admin'

export CLUSTER_HOSTNAME='localhost'

export CLUSTER_PORT='3306'

# proxysql monitoring user. proxysql admin script will create this user in pxc to monitor pxc-nodes.

export MONITOR_USERNAME="monitor"

export MONITOR_PASSWORD="monitor"

# Application user to connect to pxc-node through proxysql

export CLUSTER_APP_USERNAME="sbuser"

export CLUSTER_APP_PASSWORD="sbpass"

# ProxySQL hostgroup IDs

export WRITER_HOSTGROUP_ID='10'

export READER_HOSTGROUP_ID='11'

# ProxySQL read/write configuration mode.

export MODE="singlewrite"

# max_connections default (used only when INSERTing a new mysql_servers entry)

export MAX_CONNECTIONS="1000"

# Determines the maximum number of writesets a node can have queued

# before the node is SHUNNED to avoid stale reads.

export MAX_TRANSACTIONS_BEHIND=100

# Connections to the backend servers (from ProxySQL) will use SSL

export USE_SSL="no"

Load balancing with ProxySQL

Post installation you can login to the admin interface, use the credentials, host name and port specified in config file.

[root@standalone ~]# mysql -u admin -padmin -h 127.0.0.1 -P6032

mysql: [Warning] Using a password on the command line interface can be insecure.

Welcome to the MySQL monitor. Commands end with ; or \g.

Your MySQL connection id is 15421

Server version: 5.5.30 (ProxySQL Admin Module)

Oracle is a registered trademark of Oracle Corporation and/or its

affiliates. Other names may be trademarks of their respective

owners.

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql>

Adding Cluster Nodes to proxySQL

As mentioned in the config file we will use two hostgroup id’s for write & read

export WRITER_HOSTGROUP_ID='10' ⇒ It mentions the node is a write node

export READER_HOSTGROUP_ID='11' ⇒ It mentions the node is a read node

On proxy-node please execute the following queries,

INSERT INTO mysql_servers(hostgroup_id, hostname, port,weight,comment) VALUES (10,'192.168.56.102',3306,1000000,'WRITE');

INSERT INTO mysql_servers(hostgroup_id, hostname, port,weight,comment) VALUES (11,'192.168.56.103',3306,1000000,'READ');

INSERT INTO mysql_servers(hostgroup_id, hostname, port,weight,comment) VALUES (11,'192.168.56.107',3306,1000000,'READ');

Once added we need to load them and save to disk.

LOAD MYSQL SERVERS TO RUNTIME;
SAVE MYSQL SERVERS TO DISK;

ProxySQL Monitoring

In-order to enable monitoring of PXC nodes in ProxySQL, we will create a user on any node in the cluster.

Execute the below query on any one of PXC node.

CREATE USER 'monitor'@'%' IDENTIFIED BY 'monitor';

GRANT USAGE ON *.* TO 'monitor'@'%';

Now configure the user in ProxySQL

UPDATE global_variables SET variable_value='monitor' WHERE variable_name="mysql-monitor_username";

UPDATE global_variables SET variable_value='monitor' WHERE variable_name="mysql-monitor_password";

Once added we need to load them and save to disk

LOAD MYSQL VARIABLES TO RUNTIME;

SAVE MYSQL VARIABLES TO DISK;

To confirm that monitoring is enabled, check the monitoring logs:

mysql> SELECT * FROM monitor.mysql_server_connect_log ORDER BY time_start_us DESC LIMIT 6;

+----------------+------+------------------+-------------------------+---------------+

+----------------+------+------------------+-------------------------+---------------+

| 192.168.56.103 | 3306 | 1558139759619458 | 1956 | NULL |

| 192.168.56.102 | 3306 | 1558139758921182 | 1336 | NULL |

| 192.168.56.107 | 3306 | 1558139758223625 | 1486 | NULL |

| 192.168.56.107 | 3306 | 1558139699644073 | 1453 | NULL |

| 192.168.56.102 | 3306 | 1558139698932583 | 1580 | NULL |

| 192.168.56.103 | 3306 | 1558139698220689 | 1135 | NULL |

+----------------+------+------------------+-------------------------+---------------+

6 rows in set (0.00 sec)

mysql> SELECT * FROM monitor.mysql_server_ping_log ORDER BY time_start_us DESC LIMIT 6;

+----------------+------+------------------+----------------------+------------+

+----------------+------+------------------+----------------------+------------+

| 192.168.56.103 | 3306 | 1558139789634032 | 590 | NULL |

| 192.168.56.107 | 3306 | 1558139789500496 | 385 | NULL |

| 192.168.56.102 | 3306 | 1558139789367311 | 416 | NULL |

| 192.168.56.102 | 3306 | 1558139779544502 | 530 | NULL |

| 192.168.56.103 | 3306 | 1558139779455180 | 480 | NULL |

| 192.168.56.107 | 3306 | 1558139779366472 | 500 | NULL |

+----------------+------+------------------+----------------------+------------+

6 rows in set (0.00 sec)

Query Rules

We will now configure ProxySQL to send the read queries to read nodes and everything else to the master, also you can add what all types of select needs to be diverted.

INSERT INTO mysql_query_rules (rule_id,active,match_pattern,destination_hostgroup,apply) VALUES (1,1,'^SELECT.*FOR UPDATE$',10,1), (2,1,'^SELECT',11,1);

Once added we need to load them and save to disk

LOAD MYSQL QUERY RULES TO RUNTIME;

SAVE MYSQL QUERY RULES TO DISK;

Creating Application user

Now we will create the application user which will connect to application from ProxySQL

Execute the below query on any one of PXC node.

CREATE USER 'sbuser'@'192.168.56.106' IDENTIFIED BY 'sbpass';

GRANT ALL ON *.* TO 'sbuser'@'192.168.56.106';

Now add them to mysql_users table on ProxySQL

INSERT INTO mysql_users (username,password,default_hostgroup) VALUES ('sbuser','sbpass',10);

Once added we need to load them and save to disk

LOAD MYSQL USERS TO RUNTIME;

SAVE MYSQL USERS TO DISK;

Scheduler

Scheduler is a cron-like implementation integrated inside ProxySQL with millisecond granularity. We have a inbuilt script proxysql_galera_checker which checks the status of nodes. This tool will automatically handle the failover of a node, ie, if the write node goes down it will automatically promote a reader node as write node.

Execute the below command on proxy node

INSERT INTO scheduler (active,interval_ms,filename,arg1,comment) VALUES (1,10000,'/usr/bin/proxysql_galera_checker','--config-file=/etc/proxysql-admin.cnf --write-hg=10 --read-hg=11 --writer-count=0 --mode=singlewrite --log=/var/lib/proxysql/cluster_one_proxysql_galera_check.log','cluster_one');

Once added we need to load them and save to disk

LOAD SCHEDULER TO RUNTIME;

SAVE SCHEDULER TO DISK;

You can check the same from server,

mysql> select * from scheduler\G

*************************** 1. row ***************************

id: 1

active: 1

interval_ms: 10000

filename: /usr/bin/proxysql_galera_checker

arg1: --config-file=/etc/proxysql-admin.cnf --write-hg=10 --read-hg=11 --writer-count=0 --mode=singlewrite --log=/var/lib/proxysql/cluster_one_proxysql_galera_check.log

arg2: NULL

arg3: NULL

arg4: NULL

arg5: NULL

comment: cluster_one

1 row in set (0.00 sec)

This scheduler, as soon started will write the logs to the log file .

Read/Write benchmark test

Now we can run sysbench to know how PXC works with proxySQL, here to validate i will run sysbench for writes, reads & RW separately and will show what nodes where used for them.

Before running benchmark we need to create a schema/database on any one of the cluster node.

create database sbtest;

Now we need to create a table and load data into it.

sysbench --report-interval=5 --threads=4 --num-requests=0 --time=20 /usr/share/sysbench/tests/include/oltp_legacy/oltp.lua --mysql-user='sbuser' --mysql-password='sbpass' --oltp-table-size=10000 --mysql-host=127.0.0.1 --mysql-port=6033 prepare

[root@standalone]# sysbench --report-interval=5 --threads=4 --num-requests=0 --time=20 /usr/share/sysbench/tests/include/oltp_legacy/oltp.lua --mysql-user='sbuser' --mysql-password='sbpass' --oltp-table-size=10000 --mysql-host=127.0.0.1 --mysql-port=6033 prepare

sysbench 1.0.17 (using bundled LuaJIT 2.1.0-beta2)

Creating table 'sbtest1'...

Inserting 10000 records into 'sbtest1'

Creating secondary indexes on 'sbtest1'...

[root@standalone]#

Write test

sysbench '/usr/share/sysbench/oltp_update_index.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user='sbuser' --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

[root@standalone]# sysbench '/usr/share/sysbench/oltp_update_index.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user='sbuser' --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

sysbench 1.0.17 (using bundled LuaJIT 2.1.0-beta2)

Running the test with following options:

Number of threads: 4

Initializing random number generator from current time

Initializing worker threads...

Threads started!

SQL statistics:

queries performed:

write: 14709

other: 11914

total: 26623

transactions: 26623 (887.24 per sec.)

queries: 26623 (887.24 per sec.)

ignored errors: 0 (0.00 per sec.)

reconnects: 0 (0.00 per sec.)

General statistics:

total time: 30.0051s

total number of events: 26623

Latency (ms):

min: 0.19

avg: 4.50

max: 146.37

95th percentile: 10.46

sum: 119913.56

Threads fairness:

events (avg/stddev): 6655.7500/73.86

execution time (avg/stddev): 29.9784/0.00

[root@standalone sysbench]#

Checking stats table you can see on srv_host 192.168.56.102 check the queries count

mysql> select * from stats_mysql_connection_pool where hostgroup in (10,11) order by hostgroup,srv_host ;

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

| 10 | 192.168.56.102 | 3306 | ONLINE | 0 | 4 | 4 | 0 | 26627 | 639100 | 0 | 455 |

| 11 | 192.168.56.103 | 3306 | ONLINE | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 452 |

| 11 | 192.168.56.107 | 3306 | ONLINE | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 512 |

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

3 rows in set (0.00 sec)

Read Test

sysbench '/usr/share/sysbench/oltp_point_select.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user=sbuser --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

[root@standalone]# sysbench '/usr/share/sysbench/oltp_point_select.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user=sbuser --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

sysbench 1.0.17 (using bundled LuaJIT 2.1.0-beta2)

Running the test with following options:

Number of threads: 4

Initializing random number generator from current time

Initializing worker threads...

Threads started!

SQL statistics:

queries performed:

write: 0

other: 0

total: 259526

transactions: 259526 (8650.08 per sec.)

queries: 259526 (8650.08 per sec.)

ignored errors: 0 (0.00 per sec.)

reconnects: 0 (0.00 per sec.)

General statistics:

total time: 30.0015s

total number of events: 259526

Latency (ms):

min: 0.17

avg: 0.46

max: 20.90

95th percentile: 0.69

sum: 119449.74

Threads fairness:

events (avg/stddev): 64881.5000/3101.19

execution time (avg/stddev): 29.8624/0.01

[root@standalone sysbench]#

Checking stats table you can see on srv_host 192.168.56.103 & .107 check the queries count getting affected from 0 to new values but the node1 (192.168.56.102) remains the same. This states the read nodes are only used for read not the write node.

mysql> select * from stats_mysql_connection_pool where hostgroup in (10,11) order by hostgroup,srv_host ;

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

| 10 | 192.168.56.102 | 3306 | ONLINE | 0 | 4 | 4 | 0 | 26627 | 639100 | 0 | 408 |

| 11 | 192.168.56.103 | 3306 | ONLINE | 0 | 3 | 3 | 0 | 128722 | 3089358 | 8954316 | 425 |

| 11 | 192.168.56.107 | 3306 | ONLINE | 0 | 3 | 3 | 0 | 130810 | 3139466 | 9087246 | 393 |

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

3 rows in set (0.00 sec)

Read/Write Test

sysbench '/usr/share/sysbench/oltp_read_write.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user='sbuser' --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

[root@standalone]# sysbench '/usr/share/sysbench/oltp_read_write.lua' --table-size=1000000 --mysql-host=192.168.56.106 --mysql-db='sbtest' --mysql-user='sbuser' --mysql-password='sbpass' --mysql-port=6033 --time=30 --threads=4 --db-driver=mysql run

sysbench 1.0.17 (using bundled LuaJIT 2.1.0-beta2)

Running the test with following options:

Number of threads: 4

Initializing random number generator from current time

Initializing worker threads...

Threads started!

SQL statistics:

queries performed:

write: 17417

other: 19607

total: 123432

transactions: 6170 (205.54 per sec.)

queries: 123432 (4111.85 per sec.)

ignored errors: 2 (0.07 per sec.)

reconnects: 0 (0.00 per sec.)

General statistics:

total time: 30.0171s

total number of events: 6170

Latency (ms):

min: 8.76

avg: 19.45

max: 102.21

95th percentile: 27.17

sum: 120008.01

Threads fairness:

events (avg/stddev): 1542.5000/103.42

execution time (avg/stddev): 30.0020/0.00

[root@standalone sysbench]#

Checking stats table you can see on the difference of queries count.

mysql> select * from stats_mysql_connection_pool where hostgroup in (10,11) order by hostgroup,srv_host ;

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

| 10 | 192.168.56.102 | 3306 | ONLINE | 0 | 4 | 4 | 0 | 63677 | 3249956 | 0 | 373 |

| 11 | 192.168.56.103 | 3306 | ONLINE | 0 | 3 | 3 | 0 | 172881 | 4250574 | 82798911 | 303 |

| 11 | 192.168.56.107 | 3306 | ONLINE | 0 | 3 | 3 | 0 | 173087 | 4251066 | 79370204 | 328 |

+-----------+----------------+----------+--------+----------+----------+--------+---------+---------+-----------------+-----------------+------------+

3 rows in set (0.00 sec)

Automatic failover

If a node goes down proxysql will automatically make the node as unavailable for connections and will make it to state OFFLINE_SOFT/OFFLINE_HARD

OFFLINE_SOFT : Active transactions and connections will still be used, but no new traffic will be send to the node.

OFFLINE_HARD : All the current traffic will be immediately terminated, and no new traffic will be sent.

Scenario 1 : Making one of the read node down.

[root@galera1 ~]# hostname -I

192.168.56.103

[root@galera2 ~]# service mysql status

SUCCESS! MySQL (Percona XtraDB Cluster) running (2820)

[root@galera2 ~]# service mysql stop

Shutting down MySQL (Percona XtraDB Cluster).............. SUCCESS!

[root@galera2 ~]#

Checking the status on proxy

mysql> SELECT hostgroup_id,hostname,port,status,comment FROM mysql_servers;

+--------------+----------------+------+--------------+---------+

+--------------+----------------+------+--------------+---------+

| 10 | 192.168.56.102 | 3306 | ONLINE | WRITE |

| 11 | 192.168.56.103 | 3306 | OFFLINE_HARD | READ |

| 11 | 192.168.56.107 | 3306 | ONLINE | READ |

+--------------+----------------+------+--------------+---------+

3 rows in set (0.00 sec)

mysql>

After making the node up

mysql> SELECT hostgroup_id,hostname,port,status,comment FROM mysql_servers;

+--------------+----------------+------+--------+---------+

+--------------+----------------+------+--------+---------+

| 10 | 192.168.56.102 | 3306 | ONLINE | WRITE |

| 11 | 192.168.56.103 | 3306 | ONLINE | READ |

| 11 | 192.168.56.107 | 3306 | ONLINE | READ |

+--------------+----------------+------+--------+---------+

3 rows in set (0.00 sec)

mysql>

Scenario 2 : Making the current write node down.

[root@galera1 ~]# hostname -I

192.168.56.102

[root@galera1 ~]# service mysql status

SUCCESS! MySQL (Percona XtraDB Cluster) running (2723)

[root@galera1 ~]# service mysql stop

Shutting down MySQL (Percona XtraDB Cluster)............... SUCCESS!

[root@galera1 ~]#

Checking the status on proxy, you can see here proxy has automatically promoted another read node as write node

mysql> SELECT hostgroup_id,hostname,port,status,comment FROM mysql_servers;

+--------------+----------------+------+--------------+---------+

+--------------+----------------+------+--------------+---------+

| 11 | 192.168.56.102 | 3306 | OFFLINE_HARD | READ |

| 11 | 192.168.56.103 | 3306 | ONLINE | READ |

| 10 | 192.168.56.107 | 3306 | ONLINE | WRITE |

+--------------+----------------+------+--------------+---------+

3 rows in set (0.00 sec)

mysql>

After making the node up , the node has changed to READ node.

mysql> SELECT hostgroup_id,hostname,port,status,comment FROM mysql_servers;

+--------------+----------------+------+--------+---------+

+--------------+----------------+------+--------+---------+

| 11 | 192.168.56.102 | 3306 | ONLINE | READ |

| 11 | 192.168.56.103 | 3306 | ONLINE | READ |

| 10 | 192.168.56.107 | 3306 | ONLINE | WRITE |

+--------------+----------------+------+--------+---------+

3 rows in set (0.00 sec)

mysql>

So whenever a write nodes goes down proxySQL will automatically promote another node as write node and make sure writes won’t fail. Here failover is done automatically.

The Database Solution

Wednesday, 22 May 2019

Load balancing PXC with ProxySQL

No comments:

Translate