The first tutorial is a high level overview of how to install ProxySQL , and how to quickly configure and use it.
Install and compile ProxySQL
Installing and compiling ProxySQL is very straightforward:
1) download and uncompress the source code from GitHub
2) compile
Note:
The following packages are needed in order to compile:
- libssl and ibssl-dev (Ubuntu) or openssl and openssl-devel (Centos)
- libglib2 and libglib2.0-dev (Ubuntu) or glib2 and glib2-devel (Centos)
- libmysqlclient and libmysqlclient-dev (Ubuntu) or mysql-libs and mysql-devel (Centos)
Below an example:
vegaicm@voyager:~/proxysql$ wget -q https://github.com/renecannao/proxysql/archive/master.zip -O proxysql.zip
vegaicm@voyager:~/proxysql$ unzip -q proxysql.zip
vegaicm@voyager:~/proxysql$ cd proxysql-master/src/
vegaicm@voyager:~/proxysql/proxysql-master/src$ make
...
gcc -o proxysql obj/main.o obj/free_pkts.o obj/mem.o obj/debug.o obj/sqlite3.o obj/fundadb_hash.o obj/global_variables.o obj/mysql_connpool.o obj/mysql_protocol.o obj/mysql_handler.o obj/network.o obj/queue.o obj/threads.o obj/admin_sqlite.o -I../include -lpthread -ggdb -rdynamic -lcrypto `mysql_config --libs_r --cflags` `pkg-config --libs --cflags gthread-2.0` -DDEBUG -lm
Configure ProxySQL to run against your local MySQL instance
ProxySQL is under active development, evolving day by day thus some of the options and switches may change from version to version.
To simplify the configuration and to make it easy to try, ProxySQL is shipped with an interactive script that should help its configuration : proxysql_interactive_config.pl
Launch the script, and it will guide you step by step through a basic configuration.
The script will propose several default values, making the configuration very simple.
Following is an example, where we are going to use many of the defaults and configure only server and usernames:
vegaicm@voyager:~/proxysql/proxysql-master/src$ ./proxysql_interactive_config.pl
Interactive script to configure ProxySQL,
High Performance and High Availability proxy for MySQL
Generic options:
- core_dump_file_size : maximum size of core dump in case of crash
- stack_size : stack size allocated for each thread
- error_log : log file for error messages (not implemented yet)
core_dump_file_size [0]:
stack_size (65536-8388608) [524288]:
Clients can communicate with ProxySQL through 2 different sockets:
- proxy_mysql_port : TCP socket for MySQL traffic : default is 6033
- mysql_socket : Unix Domanin socket : default is /tmp/proxysql.sock
proxy_mysql_port [6033]:
mysql_socket [/tmp/proxysql.sock]:
ProxySQL uses an admin interface for runtime configuration and to export statistics.
Such interface uses the MySQL protocol and can be used by any MySQL client.
Options:
- proxy_admin_port : TCP socket for Administration : default is proxy_mysql_port-1 (6032)
- proxy_admin_user : username for authentication ( this is not a mysql user )
- proxy_admin_password : password for the user specified in proxy_admin_user
proxy_admin_port [6032]:
proxy_admin_user [admin]:
proxy_admin_password [admin]:
ProxySQL allows to cache SELECT statements executed by the application.
Query cache is configured through:
- mysql_query_cache_partitions : defines the number of partitions, reducing contention
- mysql_query_cache_default_timeout : defaults TTL for queries without explicit TTL
- mysql_query_cache_size : total amount of memory allocable for query cache
mysql_query_cache_partitions (1-64) [16]:
mysql_query_cache_default_timeout (0-315360000) [1]: 30
mysql_query_cache_size (1048576-10737418240) [67108864]:
Several options define the network behaviour of ProxySQL:
- mysql_threads : defines how many threads will process MySQL traffic
- mysql_poll_timeout : poll() timeout (millisecond)
- mysql_max_query_size : maximum length of a query to be analyzed
- mysql_max_resultset_size : maximum size of resultset for caching and buffering
- net_buffer_size : internal buffer for network I/O
- backlog : listen() backlog
mysql_threads (1-128) [8]: 4
mysql_poll_timeout (100-1000000) [10000]:
mysql_max_query_size (1-16777210) [1048576]:
mysql_max_resultset_size (1-1073741824) [1048576]:
net_buffer_size (1024-16777216) [8192]:
backlog (50-10000) [2000]:
ProxySQL implements an internal connection pool. Configurable with:
- mysql_connection_pool_enabled : enables the connection pool if set to 1
- mysql_wait_timeout : timeout to drop unused connections
mysql_connection_pool_enabled (0-1) [1]:
mysql_wait_timeout (1-31536000) [28800]:
ProxySQL connects to various mysqld instances that form the backend.
- mysql_servers : list of mysqld servers in the format host:port;host:port;...
Hostname[:port] of backend#1 : 127.0.0.1:3306
Would you like to add another backend server (Y-N) [N]: N
Few options specify how to connect to the backend:
- mysql_usage_user : user used by ProxySQL to connect to the backend to verify its status
- mysql_usage_password : password for user specified in mysql_usage_user
Note:
the user specified in mysql_usage_user needs only USAGE privilege, and you can create the user with GRANT USAGE
mysql_usage_user [proxy]:
mysql_usage_password [proxy]:
Note (again!):
The user specified in mysql_usage_user needs only USAGE privilege
You can create the user with GRANT USAGE ON *.* TO 'proxy'@'<proxysqlip>' IDENTIFIED BY 'proxy';
ProxySQL authenticates clients' connections, and then uses the same credentials to connect to the backends.
ProxySQL needs to know clients' usernames and passwords because a single client connection can generate multiple connections to the backend.
Username for user#1 : testuser
Password for user msandbox : testpassword
Would you like to add another user (Y-N) [N]: N
Basic configuration completed!
If you compiled ProxySQL with debug information (enabled by default) you can enable debug verbosity.
Would you like to enable debug verbosity? (Y-N) [N]: N
#
# ProxySQL config file
#
# Generated using proxysql_interactive_config.pl
#
[global]
core_dump_file_size=0
debug=0
stack_size=524288
proxy_admin_port=6032
proxy_admin_user=admin
proxy_admin_password=admin
net_buffer_size=8192
backlog=2000
[mysql]
mysql_threads=4
proxy_mysql_port=6033
mysql_socket=/tmp/proxysql.sock
mysql_query_cache_partitions=16
mysql_query_cache_default_timeout=30
mysql_query_cache_size=67108864
mysql_poll_timeout=10000
mysql_max_query_size=1048576
mysql_max_resultset_size=1048576
mysql_connection_pool_enabled=1
mysql_wait_timeout=28800
mysql_servers=127.0.0.1:3306
mysql_usage_user=proxy
mysql_usage_password=proxy
[mysql users]
testuser=testpassword
[debug]
debug_generic=0
debug_net=0
debug_pkt_array=0
debug_memory=0
debug_mysql_com=0
debug_mysql_connection=0
debug_mysql_server=0
debug_admin=0
debug_mysql_auth=0
Would you like to write a configuration file? (Y-N) [Y]:
config filename [proxysql.cnf]:
Configuration completed!
Quit
Note on debugging: do not enable debug at this stage. I will write an article just on this topic.
Create users in your local MySQL instance to allow ProxySQL to connect:
GRANT USAGE ON *.* TO 'proxy'@'127.0.0.1' IDENTIFIED BY 'proxy';
CREATE USER 'testuser'@'127.0.0.1' IDENTIFIED BY 'testpassword';
GRANT ALL PRIVILEGES ON test.* TO 'testuser'@'127.0.0.1';
GRANT ALL PRIVILEGES ON sbtest.* TO 'testuser'@'127.0.0.1';
Run ProxySQL
Starting ProxySQL is very straightforward, just execute the following in your prompt :
vegaicm@voyager:~/proxysql/proxysql-master/src$ ./proxysql
Note that currently it has some intentional limitations due the fact that is under development:
- no daemon support, only foreground;
- no logging facilities included, only stdout;
- the working directory must be writable .
Connect to MySQL through ProxySQL
Let’s make a simple test to connect to ProxySQL using tcp port 6033 :
vegaicm@voyager:~$ mysql -u testuser -ptestpassword -h 127.0.0.1 -P6033
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 3055020864
Server version: 5.1.30 (Ubuntu)
Copyright (c) 2000, 2013, 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> SHOW GRANTS;
+-----------------------------------------------------------------------------------------------------------------+
| Grants for testuser@127.0.0.1 |
+-----------------------------------------------------------------------------------------------------------------+
| GRANT USAGE ON *.* TO 'testuser'@'127.0.0.1' IDENTIFIED BY PASSWORD '*9F69E47E519D9CA02116BF5796684F7D0D45F8FA' |
| GRANT ALL PRIVILEGES ON `test`.* TO 'testuser'@'127.0.0.1' |
| GRANT ALL PRIVILEGES ON `sbtest`.* TO 'testuser'@'127.0.0.1' |
+-----------------------------------------------------------------------------------------------------------------+
3 rows in set (0.00 sec)
mysql>
mysql> use test;
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
mysql> CREATE TABLE proxytest (id INT NOT NULL AUTO_INCREMENT PRIMARY KEY) ENGINE=INNODB;
Query OK, 0 rows affected (0.04 sec)
mysql> INSERT INTO proxytest VALUES(NULL);
Query OK, 1 row affected (0.00 sec)
mysql> INSERT INTO proxytest SELECT NULL FROM proxytest;
Query OK, 1 row affected (0.01 sec)
Records: 1 Duplicates: 0 Warnings: 0
mysql> INSERT INTO proxytest SELECT NULL FROM proxytest;
Query OK, 2 rows affected (0.00 sec)
Records: 2 Duplicates: 0 Warnings: 0
mysql> SELECT * FROM proxytest;
+----+
| id |
+----+
| 1 |
| 2 |
| 3 |
| 4 |
+----+
4 rows in set (0.00 sec)
mysql> \q
Bye
Now let’s try to connect to ProxySQL through socket /tmp/proxysql.sock :
vegaicm@voyager:~$ mysql -u testuser -ptestpassword -S /tmp/proxysql.sock test
Reading table information for completion of table and column names
You can turn off this feature to get a quicker startup with -A
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 3054492480
Server version: 5.1.30 (Ubuntu)
Copyright (c) 2000, 2013, 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> SELECT * FROM proxytest;
+----+
| id |
+----+
| 1 |
| 2 |
| 3 |
| 4 |
+----+
4 rows in set (0.00 sec)
mysql> \q
Bye
Simple queries using ProxySQL seems to give us the results we were expecting.
Using mysqlslap we can quickly test how ProxySQL performs compared to MySQL running the very simple query “SELECT * FROM proxytest”. We will run MySQL with and without MySQL query cache enabled.
vegaicm@voyager:~$ for i in `seq 1 20000` ; do echo "SELECT * FROM proxytest;" ; done > /tmp/select_proxytest.sql
vegaicm@voyager:~$ mysql -u root -e "SET GLOBAL query_cache_size=0;" ; mysqlslap -u testuser -ptestpassword --create-schema=test -c 8,16,32,64,128,256 -q /tmp/select_proxytest.sql -S /var/run/mysqld/mysqld.sock | grep "Average number of seconds"
Average number of seconds to run all queries: 3.284 seconds
Average number of seconds to run all queries: 6.591 seconds
Average number of seconds to run all queries: 13.530 seconds
Average number of seconds to run all queries: 29.546 seconds
Average number of seconds to run all queries: 66.452 seconds
Average number of seconds to run all queries: 138.742 seconds
vegaicm@voyager:~$ mysql -u root -e "SET GLOBAL query_cache_size=1024*1024;" ; mysqlslap -u testuser -ptestpassword --create-schema=test -c 8,16,32,64,128,256 -q /tmp/select_proxytest.sql -S /var/run/mysqld/mysqld.sock | grep "Average number of seconds"
Average number of seconds to run all queries: 1.632 seconds
Average number of seconds to run all queries: 4.528 seconds
Average number of seconds to run all queries: 7.533 seconds
Average number of seconds to run all queries: 17.182 seconds
Average number of seconds to run all queries: 41.967 seconds
Average number of seconds to run all queries: 91.434 seconds
vegaicm@voyager:~$ mysqlslap -u testuser -ptestpassword --create-schema=test -c 8,16,32,64,128,256 -q /tmp/select_proxytest.sql -S /tmp/proxysql.sock | grep "Average number of seconds"
Average number of seconds to run all queries: 1.135 seconds
Average number of seconds to run all queries: 1.908 seconds
Average number of seconds to run all queries: 4.575 seconds
Average number of seconds to run all queries: 9.451 seconds
Average number of seconds to run all queries: 18.843 seconds
Average number of seconds to run all queries: 36.143 seconds
A picture is worth a thousands words:
Build a small cluster to run ProxySQL
While it is possible to use ProxySQL on a setup with a single MySQL server, to use many of its features (load balancing, queries routing, read/write split) it is required to have several MySQL backends, configured in various architectures.
To build a small cluster with several MySQL backends we will create a simple replication setup using MySQL Sandbox .
If you are not familiar with MySQL Sandbox, it is “a tool for installing one or more MySQL servers in isolation, without affecting other servers” that you can download from https://launchpad.net/mysql-sandbox .
vegaicm@voyager:~$ make_replication_sandbox mysql_binaries/mysql-5.5.34-linux2.6-i686.tar.gz
installing and starting master
installing slave 1
installing slave 2
starting slave 1
.... sandbox server started
starting slave 2
.... sandbox server started
initializing slave 1
initializing slave 2
replication directory installed in $HOME/sandboxes/rsandbox_mysql-5_5_34
installing and starting master
installing slave 1
installing slave 2
starting slave 1
.... sandbox server started
starting slave 2
.... sandbox server started
initializing slave 1
initializing slave 2
replication directory installed in $HOME/sandboxes/rsandbox_mysql-5_5_34
Now that the cluster is installed, verify on which ports are listening the various mysqld processes:
vegaicm@voyager:~$ cd sandboxes/rsandbox_mysql-5_5_34
vegaicm@voyager:~/sandboxes/rsandbox_mysql-5_5_34$ cat default_connection.json
{
"master":
{
"host": "127.0.0.1",
"port": "23389",
"socket": "/tmp/mysql_sandbox23389.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
,
"node1":
{
"host": "127.0.0.1",
"port": "23390",
"socket": "/tmp/mysql_sandbox23390.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
,
"node2":
{
"host": "127.0.0.1",
"port": "23391",
"socket": "/tmp/mysql_sandbox23391.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
}
vegaicm@voyager:~/sandboxes/rsandbox_mysql-5_5_34$ cat default_connection.json
{
"master":
{
"host": "127.0.0.1",
"port": "23389",
"socket": "/tmp/mysql_sandbox23389.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
,
"node1":
{
"host": "127.0.0.1",
"port": "23390",
"socket": "/tmp/mysql_sandbox23390.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
,
"node2":
{
"host": "127.0.0.1",
"port": "23391",
"socket": "/tmp/mysql_sandbox23391.sock",
"username": "msandbox@127.%",
"password": "msandbox"
}
}
The mysqld processes are listening on port 23389 (master) and 23390 and 23391 (slaves).
Important note:
Unless configured differently (details in another tutorial), ProxySQL considers a writable master any server that is configured with read_only=OFF . Because MySQL Sandbox doesn’t configure the slaves with read_only=ON , ProxySQL will consider any node as a master : that is obviously incorrect and needs to be addressed before starting ProxySQL.
Change read_only setting on the two slaves:
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P23390 -e "SET GLOBAL read_only=ON"
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P23391 -e "SET GLOBAL read_only=ON"
Reconfigure ProxySQL to use the test cluster
It is possible to reconfigure ProxySQL without restarting it, but it is a long process that will be explained in another tutorial . For now, in order to reconfigure proxysql , you should :
- stop proxysql
- delete the proxysql.db* files
- edit proxysql.cnf or relaunch proxysql_interactive_config.pl
Note: proxysql.db* files are the database files that ProxySQL uses to store configurations after processing proxysql.cnf , and the configuration stored in these files have higher priority over the same configuration values stored in proxysql.cnf . That is the reason why you should delete proxysql.db* files if you edit proxysql.cnf , or the new settings in proxysql.cnf won’t have any effect.
In the next article we will describe the content of these database files and how to reconfigure ProxySQL without deleting any file and without restarting the process.
For semplicity I will rerun proxysql_interactive_config.pl and highlight the few differences:
vegaicm@voyager:~/proxysql/proxysql-master/src$ ./proxysql
^C
vegaicm@voyager:~/proxysql/proxysql-master/src$ rm -f proxysql.db*
vegaicm@voyager:~/proxysql/proxysql-master/src$ rm proxysql.cnf
vegaicm@voyager:~/proxysql/proxysql-master/src$ ./proxysql_interactive_config.pl
Interactive script to configure ProxySQL,
High Performance and High Availability proxy for MySQL
…
...
Hostname[:port] of backend#1 : 127.0.0.1:23389
Would you like to add another backend server (Y-N) [N]: Y
Hostname[:port] of backend#2 : 127.0.0.1:23390
Would you like to add another backend server (Y-N) [N]: y
Hostname[:port] of backend#3 : 127.0.0.1:23391
Would you like to add another backend server (Y-N) [N]: N
…
…
mysql_usage_user [proxy]: msandbox
mysql_usage_password [proxy]: msandbox
Note (again!):
The user specified in mysql_usage_user needs only USAGE privilege
You can create the user with GRANT USAGE ON *.* TO 'msandbox'@'<proxysqlip>' IDENTIFIED BY 'msandbox';
ProxySQL authenticates clients' connections, and then uses the same credentials to connect to the backends.
ProxySQL needs to know clients' usernames and passwords because a single client connection can generate multiple connections to the backend.
Username for user#1 : msandbox
Password for user msandbox : msandbox
Would you like to add another user (Y-N) [N]: N
…
…
Connect to the various MySQL backends through ProxySQL
Restart ProxySQL :
vegaicm@voyager:~/proxysql/proxysql-master/src$ ./proxysql
Once again , let’s make a simple test to connect to ProxySQL :
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P6033
Welcome to the MySQL monitor. Commands end with ; or \g.
Your MySQL connection id is 3048201024
Server version: 5.1.30 MySQL Community Server (GPL)
Copyright (c) 2000, 2013, 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> SELECT 1;
+---+
| 1 |
+---+
| 1 |
+---+
1 row in set (0.00 sec)
mysql> \q
All looks good so far. But to which backend are we connected?
Let make a simple test:
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port"
+------------+--------+
| @@hostname | @@port |
+------------+--------+
| voyager | 23390 |
+------------+--------+
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port"
+------------+--------+
| @@hostname | @@port |
+------------+--------+
| voyager | 23391 |
+------------+--------+
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port"
+------------+--------+
| @@hostname | @@port |
+------------+--------+
| voyager | 23390 |
+------------+--------+
vegaicm@voyager:~$ mysql -u msandbox -pmsandbox -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port"
+------------+--------+
| @@hostname | @@port |
+------------+--------+
| voyager | 23389 |
+------------+--------+
It seems we are randomly connected to one of the backends.
Let make a different test now: we will send the same identical query over a very small period of time:
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23389
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23390
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23390
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23389
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23389
vegaicm@voyager:~$ for i in `seq 1 50` ; do mysql -u msandbox -pmsandbox -B -N -h 127.0.0.1 -P6033 -e "SELECT @@hostname, @@port" ; done | sort | uniq -c
50 voyager 23391
Caching in action!
Because ProxySQL is configured to cache every SELECT statement for 1 second, if we send several identical queries on a short period of time the queries won’t be executed by the MySQL backend but resolved directly from ProxySQL internal cache!
Running some simple load test using ProxySQL
We will use sysbench to run some simple load test.
Note that the following results should not be considered a meaningful benchmark as all the servers run on the same physical box (my laptop) and as such they don’t really scale-out. To scale-out and improve performance you need to add more servers.
Therefore, you should consider the following only as an example of how to use proxysql.
First, prepare the table:
vegaicm@voyager:~$ sysbench --test=oltp --oltp-table-size=10000 --mysql-user=msandbox --mysql-password=msandbox --mysql-db=test --mysql-host=127.0.0.1 --mysql-port=6033 --oltp-table-size=10000 prepare
sysbench 0.4.12: multi-threaded system evaluation benchmark
No DB drivers specified, using mysql
Creating table 'sbtest'...
Creating 10000 records in table 'sbtest'...
Before running benchmarks note that :
1) performance of ProxySQL are affected by the fact that the TCP protocol is really expensive;
2) by default, sysbench uses the MySQL binary protocol ( that means it uses prepared statements ), but ProxySQL doesn’t recognize (yet?) this protocol, therefore can’t boost performance if prepared statements are used. Note that the majority of the applications do not use prepared statements, therefore ProxySQL is meant to boost the performance of the majority of the applications out there.
With this in mind, we will run some read only workload using Unix domain socket and disabling prepared statement in sysbench.
Running sysbench directly against mysqld without ProxySQL:
vegaicm@voyager:~$ sysbench --num-threads=16 --max-requests=0 --max-time=30 --test=oltp --oltp-table-size=10000 --mysql-user=msandbox --mysql-password=msandbox --mysql-db=test --mysql-socket=/tmp/mysql_sandbox23389.sock --oltp-table-size=10000 --oltp-read-only --db-ps-mode=disable run | grep 'transactions:'
transactions: 39824 (1327.14 per sec.)
And with ProxySQL:
vegaicm@voyager:~$ sysbench --num-threads=16 --max-requests=0 --max-time=30 --test=oltp --oltp-table-size=10000 --mysql-user=msandbox --mysql-password=msandbox --mysql-db=test --mysql-socket=/tmp/proxysql.sock --oltp-table-size=10000 --oltp-read-only --db-ps-mode=disable run | grep 'transactions:'
transactions: 80899 (2696.14 per sec.)
You can easily note how ProxySQL improves performance!
To summarize in a graph:
Testing read/write workload
What about using ProxySQL in a read/write workload?
sysbench with a read/write workload without proxysql:
vegaicm@voyager:~$ sysbench --num-threads=16 --max-requests=0 --max-time=30 --test=oltp --oltp-table-size=10000 --mysql-user=msandbox --mysql-password=msandbox --mysql-db=test --mysql-socket=/tmp/mysql_sandbox23389.sock --oltp-table-size=10000 --db-ps-mode=disable run | grep 'transactions:'
transactions: 13010 (433.45 per sec.)
sysbench with a read/write workload with proxysql:
vegaicm@voyager:~$ sysbench --num-threads=16 --max-requests=0 --max-time=30 --test=oltp --oltp-table-size=10000 --mysql-user=msandbox --mysql-password=msandbox --mysql-db=test --mysql-socket=/tmp/proxysql.sock --oltp-table-size=10000 --db-ps-mode=disable run | grep 'transactions:'
transactions: 14696 (487.60 per sec.)
The above shows that proxysql boosts performance also in read/write workload, but what it is more important is that transparently connects the clients to the correct servers, whatever is the master or a slave.
As mentioned previously, the above results are only an example of how to use proxysql : in another article we will run real benchmark adding more servers and highlighting how ProxySQL is able to scale-out the performance of your application without code changes.
In the next article we will discuss some of the internals of ProxySQL, and how to configure them.
I would like to see "prepared statements" support implemented. Keep in mind that in real-world apps that values security, no sane developer writes and executes raw sql with non-bounded parameters.
ReplyDeleteGetting: /root/proxysql-master/src/main.c:146: undefined reference to `g_get_monotonic_time'
ReplyDeletewhen running make. This is on centos 6.4 x64 (fresh install)
Thank you for the update. Everything compiles and runs fine on Centos now. Is there more documentation on tuning variables? I have the following happening:
ReplyDeleteI have a sphinx instance running that sends a query to proxySQL that should pull back all of the records from a mysql table. After the query has been running for a few minutes, prosySQL stops sending data, but the proxySQL process is consuming 100% cpu. I'm sure it is a matter of tweaking some variables in the config, but I have no idea where to start.
I also ran into a similar issue. Seems like a tight loop somewhere in the program. Not very stable and just has too many random issues from time to time, and the worst part is there is very little diagnostics or logs so you can't even figure out what went wrong. I have a few core dumps as well but they are all minidumps because of memory leaks not leaving enough RAM on the VM to take a full dump (on a 16GB web server with 32GB swap, MemSQL was consuming over 18GB and running into swap which certainly sounds like a memory leak). Also had an issue where one client connection was receiving data meant for a completely different client connection.
Delete@Anonymous:
DeletePlease try the Hebe branch (https://github.com/renecannao/proxysql/tree/Hebe), that handles memory usage a lot better.
If the problem isn't solved in Hebe branch, please open an issue on https://github.com/renecannao/proxysql/issues I will make sure I follow up on this issue.
Thanks.
This only appears to happen on large queries. I can index a 400MB mysql table with no problem, but our larger 3+GB tables always hang at exactly 1829MB. This doesn't happen when connecting directly to mysql.
ReplyDelete