Load balancing with ProxySQL

ProxySQL is a high-performance SQL proxy. ProxySQL runs as a daemon watched by a monitoring process. The process monitors the daemon and restarts it in case of a crash to minimize downtime.

The daemon accepts incoming traffic from MySQL clients and forwards it to backend MySQL servers.

The proxy is designed to run continuously without needing to be restarted. Most configuration can be done at runtime using queries similar to SQL statements. These include runtime parameters, server grouping, and traffic-related settings.

Note

For more information about ProxySQL, see ProxySQL documentation.

ProxySQL is available from the Percona software repositories in two versions. ProxySQL v1 does not natively support Percona XtraDB Cluster and requires custom bash scripts to keep track of the status of Percona XtraDB Cluster nodes using the ProxySQL scheduler.

ProxySQL v2 natively supports Percona XtraDB Cluster. With this version, proxysql-admin tool does not require custom scripts to keep track of Percona XtraDB Cluster status.

Manual Configuration

This tutorial describes how to configure ProxySQL with three Percona XtraDB Cluster nodes.

Node Host Name IP address
Node 1 pxc1 192.168.70.61
Node 2 pxc2 192.168.70.62
Node 3 pxc3 192.168.70.63
Node 4 proxysql 192.168.70.64

ProxySQL can be configured either using the /etc/proxysql.cnf file or through the admin interface. Using the admin interface is preferable, because it allows you to change the configuration dynamically (without having to restart the proxy).

To connect to the ProxySQL admin interface, you need a mysql client. You can either connect to the admin interface from Percona XtraDB Cluster nodes that already have the mysql client installed (Node 1, Node 2, Node 3) or install the client on Node 4 and connect locally. For this tutorial, install Percona XtraDB Cluster on Node 4:

  • On Debian or Ubuntu:

    root@proxysql:~# apt-get install percona-xtradb-cluster-client-5.7
    
  • On Red Hat Enterprise Linux or CentOS:

    [root@proxysql ~]# yum install Percona-XtraDB-Cluster-client-57
    

To connect to the admin interface, use the credentials, host name and port specified in the global variables.

Warning

Do not use default credentials in production!

The following example shows how to connect to the ProxySQL admin interface with default credentials:

root@proxysql:~# mysql -u admin -padmin -h 127.0.0.1 -P 6032

Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 2
Server version: 5.1.30 (ProxySQL Admin Module)

Copyright (c) 2009-2016 Percona LLC and/or its affiliates
Copyright (c) 2000, 2016, Oracle and/or its affiliates. All rights reserved.

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@proxysql>

To see the ProxySQL databases and tables use the following commands:

mysql@proxysql> SHOW DATABASES;
+-----+---------+-------------------------------+
| seq | name    | file                          |
+-----+---------+-------------------------------+
| 0   | main    |                               |
| 2   | disk    | /var/lib/proxysql/proxysql.db |
| 3   | stats   |                               |
| 4   | monitor |                               |
+-----+---------+-------------------------------+
4 rows in set (0.00 sec)
mysql@proxysql> SHOW TABLES;
+--------------------------------------+
| tables                               |
+--------------------------------------+
| global_variables                     |
| mysql_collations                     |
| mysql_query_rules                    |
| mysql_replication_hostgroups         |
| mysql_servers                        |
| mysql_users                          |
| runtime_global_variables             |
| runtime_mysql_query_rules            |
| runtime_mysql_replication_hostgroups |
| runtime_mysql_servers                |
| runtime_scheduler                    |
| scheduler                            |
+--------------------------------------+
12 rows in set (0.00 sec)

For more information about admin databases and tables, see Admin Tables

Note

ProxySQL has 3 areas where the configuration can reside:

  • MEMORY (your current working place)
  • RUNTIME (the production settings)
  • DISK (durable configuration, saved inside an SQLITE database)

When you change a parameter, you change it in MEMORY area. That is done by design to allow you to test the changes before pushing to production (RUNTIME), or save them to disk.

Adding cluster nodes to ProxySQL

To configure the backend Percona XtraDB Cluster nodes in ProxySQL, insert corresponding records into the mysql_servers table.

Note

ProxySQL uses the concept of hostgroups to group cluster nodes. This enables you to balance the load in a cluster by routing different types of traffic to different groups. There are many ways you can configure hostgroups (for example master and slaves, read and write load, etc.) and a every node can be a member of multiple hostgroups.

This example adds three Percona XtraDB Cluster nodes to the default hostgroup (0), which receives both write and read traffic:

mysql@proxysql> INSERT INTO mysql_servers(hostgroup_id, hostname, port) VALUES (0,'192.168.70.61',3306);
mysql@proxysql> INSERT INTO mysql_servers(hostgroup_id, hostname, port) VALUES (0,'192.168.70.62',3306);
mysql@proxysql> INSERT INTO mysql_servers(hostgroup_id, hostname, port) VALUES (0,'192.168.70.63',3306);

To see the nodes:

mysql@proxysql> SELECT * FROM mysql_servers;

+--------------+---------------+------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| hostgroup_id | hostname      | port | status | weight | compression | max_connections | max_replication_lag | use_ssl | max_latency_ms | comment |
+--------------+---------------+------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
| 0            | 192.168.70.61 | 3306 | ONLINE | 1      | 0           | 1000            | 0                   | 0       | 0              |         |
| 0            | 192.168.70.62 | 3306 | ONLINE | 1      | 0           | 1000            | 0                   | 0       | 0              |         |
| 0            | 192.168.70.63 | 3306 | ONLINE | 1      | 0           | 1000            | 0                   | 0       | 0              |         |
+--------------+---------------+------+--------+--------+-------------+-----------------+---------------------+---------+----------------+---------+
3 rows in set (0.00 sec)

Creating ProxySQL Monitoring User

To enable monitoring of Percona XtraDB Cluster nodes in ProxySQL, create a user with USAGE privilege on any node in the cluster and configure the user in ProxySQL.

The following example shows how to add a monitoring user on Node 2:

mysql@pxc2> CREATE USER 'proxysql'@'%' IDENTIFIED BY 'ProxySQLPa55';
mysql@pxc2> GRANT USAGE ON *.* TO 'proxysql'@'%';

The following example shows how to configure this user on the ProxySQL node:

mysql@proxysql> UPDATE global_variables SET variable_value='proxysql'
              WHERE variable_name='mysql-monitor_username';
mysql@proxysql> UPDATE global_variables SET variable_value='ProxySQLPa55'
              WHERE variable_name='mysql-monitor_password';

To load this configuration at runtime, issue a LOAD command. To save these changes to disk (ensuring that they persist after ProxySQL shuts down), issue a SAVE command.

mysql@proxysql> LOAD MYSQL VARIABLES TO RUNTIME;
mysql@proxysql> SAVE MYSQL VARIABLES TO DISK;

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

mysql@proxysql> SELECT * FROM monitor.mysql_server_connect_log ORDER BY time_start_us DESC LIMIT 6;
+---------------+------+------------------+----------------------+---------------+
| hostname      | port | time_start_us    | connect_success_time | connect_error |
+---------------+------+------------------+----------------------+---------------+
| 192.168.70.61 | 3306 | 1469635762434625 | 1695                 | NULL          |
| 192.168.70.62 | 3306 | 1469635762434625 | 1779                 | NULL          |
| 192.168.70.63 | 3306 | 1469635762434625 | 1627                 | NULL          |
| 192.168.70.61 | 3306 | 1469635642434517 | 1557                 | NULL          |
| 192.168.70.62 | 3306 | 1469635642434517 | 2737                 | NULL          |
| 192.168.70.63 | 3306 | 1469635642434517 | 1447                 | NULL          |
+---------------+------+------------------+----------------------+---------------+
6 rows in set (0.00 sec)
mysql> SELECT * FROM monitor.mysql_server_ping_log ORDER BY time_start_us DESC LIMIT 6;
+---------------+------+------------------+-------------------+------------+
| hostname      | port | time_start_us    | ping_success_time | ping_error |
+---------------+------+------------------+-------------------+------------+
| 192.168.70.61 | 3306 | 1469635762416190 | 948               | NULL       |
| 192.168.70.62 | 3306 | 1469635762416190 | 803               | NULL       |
| 192.168.70.63 | 3306 | 1469635762416190 | 711               | NULL       |
| 192.168.70.61 | 3306 | 1469635702416062 | 783               | NULL       |
| 192.168.70.62 | 3306 | 1469635702416062 | 631               | NULL       |
| 192.168.70.63 | 3306 | 1469635702416062 | 542               | NULL       |
+---------------+------+------------------+-------------------+------------+
6 rows in set (0.00 sec)

The previous examples show that ProxySQL is able to connect and ping the nodes you added.

To enable monitoring of these nodes, load them at runtime:

mysql@proxysql> LOAD MYSQL SERVERS TO RUNTIME;

Creating ProxySQL Client User

ProxySQL must have users that can access backend nodes to manage connections.

To add a user, insert credentials into mysql_users table:

mysql@proxysql> INSERT INTO mysql_users (username,password) VALUES ('sbuser','sbpass');
Query OK, 1 row affected (0.00 sec)

Note

ProxySQL currently doesn’t encrypt passwords.

Load the user into runtime space and save these changes to disk (ensuring that they persist after ProxySQL shuts down):

mysql@proxysql> LOAD MYSQL USERS TO RUNTIME;
mysql@proxysql> SAVE MYSQL USERS TO DISK;

To confirm that the user has been set up correctly, you can try to log in:

root@proxysql:~# mysql -u sbuser -psbpass -h 127.0.0.1 -P 6033

Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 1491
Server version: 5.1.30 (ProxySQL)

Copyright (c) 2009-2016 Percona LLC and/or its affiliates
Copyright (c) 2000, 2016, Oracle and/or its affiliates. All rights reserved.

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.

To provide read/write access to the cluster for ProxySQL, add this user on one of the Percona XtraDB Cluster nodes:

mysql@pxc3> CREATE USER 'sbuser'@'192.168.70.64' IDENTIFIED BY 'sbpass';
Query OK, 0 rows affected (0.01 sec)

mysql@pxc3> GRANT ALL ON *.* TO 'sbuser'@'192.168.70.64';
Query OK, 0 rows affected (0.00 sec)

Adding Galera Support in ProxySQL v1

ProxySQL v2 supports monitoring the status Percona XtraDB Cluster nodes. ProxySQL v1 cannot detect a node which is not in Synced state. To monitor the status of Percona XtraDB Cluster nodes in ProxySQL v1, use the proxysql_galera_checker script. The script is located here: /usr/bin/proxysql_galera_checker.

To use this script, load it into ProxySQL v1 Scheduler.

The following example shows how you can load the script for default ProxySQL v1 configuration:

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=1 --mode=singlewrite
--priority=192.168.100.20:3306,192.168.100.40:3306,192.168.100.10:3306,192.168.100.30:3306
--log=/var/lib/proxysql/cluster_one_proxysql_galera_check.log','cluster_one');

This scheduler script accepts the following options in the arg1 argument:

Option Name Required Description
--config-file Configuration File Yes Specify proxysql-admin configuration file.
--write-hg HOSTGROUP WRITERS No Specify ProxySQL write hostgroup.
--read-hg HOSTGROUP READERS No Specify ProxySQL read hostgroup.
--writer-count NUMBER WRITERS No Specify write nodes count. 0 for loadbal mode and 1 for singlewrite mode.
--mode MODE No Specify ProxySQL read/write configuration mode.
--priority WRITER PRIORITY No Specify write nodes priority.
--log LOG FILE No Specify proxysql_galera_checker log file.

Note

Specify cluster name in comment column.

To load the scheduler changes into the runtime space:

mysql@proxysql> LOAD SCHEDULER TO RUNTIME;

To make sure that the script has been loaded, check the runtime_scheduler table:

mysql@proxysql> SELECT * FROM scheduler\G
*************************** 1. row ***************************
         id: 1
     active: 1
interval_ms: 10000
   filename: /bin/proxysql_galera_checker
       arg1: --config-file=/etc/proxysql-admin.cnf --write-hg=10 --read-hg=11
             --writer-count=1 --mode=singlewrite
             --priority=192.168.100.20:3306,192.168.100.40:3306,192.168.100.10:3306,192.168.100.30:3306
             --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)

To check the status of available nodes, run the following command:

mysql@proxysql> SELECT hostgroup_id,hostname,port,status FROM mysql_servers;
+--------------+---------------+------+--------+
| hostgroup_id | hostname      | port | status |
+--------------+---------------+------+--------+
| 0            | 192.168.70.61 | 3306 | ONLINE |
| 0            | 192.168.70.62 | 3306 | ONLINE |
| 0            | 192.168.70.63 | 3306 | ONLINE |
+--------------+---------------+------+--------+
3 rows in set (0.00 sec)

Note

Each node can have the following status:

  • ONLINE: backend node is fully operational.
  • SHUNNED: backend node is temporarily taken out of use, because either too many connection errors hapenned in a short time, or replication lag exceeded the allowed threshold.
  • OFFLINE_SOFT: new incoming connections aren’t accepted, while existing connections are kept until they become inactive. In other words, connections are kept in use until the current transaction is completed. This allows to gracefully detach a backend node.
  • OFFLINE_HARD: existing connections are dropped, and new incoming connections aren’t accepted. This is equivalent to deleting the node from a hostgroup, or temporarily taking it out of the hostgroup for maintenance.

Testing Cluster with sysbench

You can install sysbench from Percona software repositories:

  • For Debian or Ubuntu:

    root@proxysql:~# apt-get install sysbench
    
  • For Red Hat Enterprise Linux or CentOS

    [root@proxysql ~]# yum install sysbench
    

Note

sysbench requires ProxySQL client user credentials that you creted in Creating ProxySQL Client User.

  1. Create the database that will be used for testing on one of the Percona XtraDB Cluster nodes:

    mysql@pxc1> CREATE DATABASE sbtest;
    
  2. Populate the table with data for the benchmark on the ProxySQL node:

    root@proxysql:~# sysbench --report-interval=5 --num-threads=4 \
    --num-requests=0 --max-time=20 \
    --test=/usr/share/doc/sysbench/tests/db/oltp.lua \
    --mysql-user='sbuser' --mysql-password='sbpass' \
    --oltp-table-size=10000 --mysql-host=127.0.0.1 --mysql-port=6033 \
    prepare
    
  3. Run the benchmark on the ProxySQL node:

    root@proxysql:~# sysbench --report-interval=5 --num-threads=4 \
      --num-requests=0 --max-time=20 \
      --test=/usr/share/doc/sysbench/tests/db/oltp.lua \
      --mysql-user='sbuser' --mysql-password='sbpass' \
      --oltp-table-size=10000 --mysql-host=127.0.0.1 --mysql-port=6033 \
      run
    

ProxySQL stores collected data in the stats schema:

mysql@proxysql> SHOW TABLES FROM stats;
+--------------------------------+
| tables                         |
+--------------------------------+
| stats_mysql_query_rules        |
| stats_mysql_commands_counters  |
| stats_mysql_processlist        |
| stats_mysql_connection_pool    |
| stats_mysql_query_digest       |
| stats_mysql_query_digest_reset |
| stats_mysql_global             |
+--------------------------------+

For example, to see the number of commands that run on the cluster:

mysql@proxysql> SELECT * FROM stats_mysql_commands_counters;
+-------------------+---------------+-----------+-----------+-----------+---------+---------+----------+----------+-----------+-----------+--------+--------+---------+----------+
| Command           | Total_Time_us | Total_cnt | cnt_100us | cnt_500us | cnt_1ms | cnt_5ms | cnt_10ms | cnt_50ms | cnt_100ms | cnt_500ms | cnt_1s | cnt_5s | cnt_10s | cnt_INFs |
+-------------------+---------------+-----------+-----------+-----------+---------+---------+----------+----------+-----------+-----------+--------+--------+---------+----------+
| ALTER_TABLE       | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| ANALYZE_TABLE     | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| BEGIN             | 2212625       | 3686      | 55        | 2162      | 899     | 569     | 1        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| CHANGE_MASTER     | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| COMMIT            | 21522591      | 3628      | 0         | 0         | 0       | 1765    | 1590     | 272      | 1         | 0         | 0      | 0      | 0       | 0        |
| CREATE_DATABASE   | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| CREATE_INDEX      | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
...
| DELETE            | 2904130       | 3670      | 35        | 1546      | 1346    | 723     | 19       | 1        | 0         | 0         | 0      | 0      | 0       | 0        |
| DESCRIBE          | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
...
| INSERT            | 19531649      | 3660      | 39        | 1588      | 1292    | 723     | 12       | 2        | 0         | 1         | 0      | 1      | 2       | 0        |
...
| SELECT            | 35049794      | 51605     | 501       | 26180     | 16606   | 8241    | 70       | 3        | 4         | 0         | 0      | 0      | 0       | 0        |
| SELECT_FOR_UPDATE | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
...
| UPDATE            | 6402302       | 7367      | 75        | 2503      | 3020    | 1743    | 23       | 3        | 0         | 0         | 0      | 0      | 0       | 0        |
| USE               | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
| SHOW              | 19691         | 2         | 0         | 0         | 0       | 0       | 1        | 1        | 0         | 0         | 0      | 0      | 0       | 0        |
| UNKNOWN           | 0             | 0         | 0         | 0         | 0       | 0       | 0        | 0        | 0         | 0         | 0      | 0      | 0       | 0        |
+-------------------+---------------+-----------+-----------+-----------+---------+---------+----------+----------+-----------+-----------+--------+--------+---------+----------+
45 rows in set (0.00 sec)

Automatic Fail-over

ProxySQL will automatically detect if a node is not available or not synced with the cluster.

You can check the status of all available nodes by running:

mysql@proxysql> SELECT hostgroup_id,hostname,port,status FROM mysql_servers;
+--------------+---------------+------+--------+
| hostgroup_id | hostname      | port | status |
+--------------+---------------+------+--------+
| 0            | 192.168.70.61 | 3306 | ONLINE |
| 0            | 192.168.70.62 | 3306 | ONLINE |
| 0            | 192.168.70.63 | 3306 | ONLINE |
+--------------+---------------+------+--------+
3 rows in set (0.00 sec)

To test problem detection and fail-over mechanism, shut down Node 3:

root@pxc3:~# service mysql stop

ProxySQL will detect that the node is down and update its status to OFFLINE_SOFT:

mysql@proxysql> SELECT hostgroup_id,hostname,port,status FROM mysql_servers;
+--------------+---------------+------+--------------+
| hostgroup_id | hostname      | port | status       |
+--------------+---------------+------+--------------+
| 0            | 192.168.70.61 | 3306 | ONLINE       |
| 0            | 192.168.70.62 | 3306 | ONLINE       |
| 0            | 192.168.70.63 | 3306 | OFFLINE_SOFT |
+--------------+---------------+------+--------------+
3 rows in set (0.00 sec)

Now start Node 3 again:

root@pxc3:~# service mysql start

The script will detect the change and mark the node as ONLINE:

mysql@proxysql> SELECT hostgroup_id,hostname,port,status FROM mysql_servers;
+--------------+---------------+------+--------+
| hostgroup_id | hostname      | port | status |
+--------------+---------------+------+--------+
| 0            | 192.168.70.61 | 3306 | ONLINE |
| 0            | 192.168.70.62 | 3306 | ONLINE |
| 0            | 192.168.70.63 | 3306 | ONLINE |
+--------------+---------------+------+--------+
3 rows in set (0.00 sec)

Assisted Maintenance Mode

Usually, to take a node down for maintenance, you need to identify that node, update its status in ProxySQL to OFFLINE_SOFT, wait for ProxySQL to divert traffic from this node, and then initiate the shutdown or perform maintenance tasks. Percona XtraDB Cluster includes a special maintenance mode for nodes that enables you to take a node down without adjusting ProxySQL manually. The mode is controlled using the pxc_maint_mode variable, which is monitored by ProxySQL and can be set to one of the following values:

  • DISABLED: This is the default state that tells ProxySQL to route traffic to the node as usual.

  • SHUTDOWN: This state is set automatically when you initiate node shutdown.

    You may need to shut down a node when upgrading the OS, adding resources, changing hardware parts, relocating the server, etc.

    When you initiate node shutdown, Percona XtraDB Cluster does not send the signal immediately. Intead, it changes the state to pxc_maint_mode=SHUTDOWN and waits for a predefined period (10 seconds by default). When ProxySQL detects that the mode is set to SHUTDOWN, it changes the status of this node to OFFLINE_SOFT, which stops creation of new connections for the node. After the transition period, any long-running transactions that are still active are aborted.

  • MAINTENANCE: You can change to this state if you need to perform maintenace on a node without shutting it down.

    You may need to isolate the node for some time, so that it does not receive traffic from ProxySQL while you resize the buffer pool, truncate the undo log, defragment or check disks, etc.

    To do this, manually set pxc_maint_mode=MAINTENANCE. Control is not returned to the user for a predefined period (10 seconds by default). When ProxySQL detects that the mode is set to MAINTENANCE, it stops routing traffic to the node. Once control is returned, you can perform maintenance activity.

    Note

    Any data changes will still be replicated across the cluster.

    After you finish maintenance, set the mode back to DISABLED. When ProxySQL detects this, it starts routing traffic to the node again.

You can increase the transition period using the pxc_maint_transition_period variable to accomodate for long-running transactions. If the period is long enough for all transactions to finish, there should hardly be any disruption in cluster workload.

During the transition period, the node continues to receive existing write-set replication traffic, ProxySQL avoids openning new connections and starting transactions, but the user can still open conenctions to monitor status.

Note

If you increase the transition period, the packaging script may determine it as a server stall.