Home > Article > Backend Development > PHP multi-process insert data
PHP multi-process insert data
- 巴扎黑Original
- 2018-05-26 09:45:023339browse
In the virtual machine centos7, single core, 1G memory
/**
* 模拟并发请求，10万次写入数据库
* 拆分为10个进程，每个进程处理一万条插入
*/
$total = 10000;
$num = 10;
$per = $total/$num;
$sql = '';
$child = '';
echo 'start '.microtime(true).PHP_EOL;
for($i = 1; $i<= $num; $i++)
{
$pid = pcntl_fork();
if($pid == -1) {
die('fork error');
}
if($pid > 0) {
//$id = pcntl_wait($status,WNOHANG);
$child[] = $pid;
} else if ($pid == 0) {
$link = mysqli_connect('localhost','root','root','yii2advanced');
$start = ($i-1)*$per + 1;
$end = $start + $per;
for($j = $start; $j< $end; $j++){
$time = microtime(true);
$sql = 'insert pcntl_test (rank,time) values ('.$j.','.$time.')';
mysqli_query($link,$sql);
}
mysqli_close($link);
$id = getmypid();
echo 'child '.$id.' finished '.microtime(true).PHP_EOL;
exit(0);
}
}
while(count($child)){
foreach($child as $k => $pid) {
$res = pcntl_waitpid($pid, $status, WNOHANG);
if ( -1 == $res || $res > 0) {
unset($child[$k]);
}
}
}
echo 'end '.microtime(true).PHP_EOL;
When $total=10000, $num = 10; the execution result is as follows:
start 1491903371.5548 child 19860 finished 1491903417.2113 child 19857 finished 1491903417.6909 child 19864 finished 1491903417.7793 child 19855 finished 1491903417.8695 child 19859 finished 1491903417.9162 child 19861 finished 1491903418.0089 child 19856 finished 1491903418.0532 child 19863 finished 1491903418.0842 child 19862 finished 1491903418.1474 child 19858 finished 1491903418.4341 end 1491903418.4424 总时间为46.88759994506836秒
When $total=10000,$ When num = 100, the execution result is as follows:
start 1491904334.1735 child 20085 finished 1491904337.0712 child 20086 finished 1491904337.144 …… child 20262 finished 1491904341.5602 child 20264 finished 1491904341.5803 end 1491904341.5869 总时间为7.413399934768677
When $total=10000, $num = 1000, the execution result is as follows:
start 1491904562.0166 child 20282 finished 1491904562.1191 child 20277 finished 1491904562.1268 child 20279 finished 1491904562.1352 ... child 21586 finished 1491904576.6954 child 21582 finished 1491904576.7024 child 21584 finished 1491904576.7226 end 1491904576.7297 总时间为14.71310019493103，相比100个子进程，耗时更长了。进程切换太多，影响了了效率应该是原因之一。
When $total=100000, $num=100 , 100,000 records, 100 processes inserting
start 1491905670.2652 child 21647 finished 1491905725.4382 child 21651 finished 1491905725.4595 child 21642 finished 1491905725.5402 .... child 21810 finished 1491905729.7709 child 21812 finished 1491905729.8498 child 21811 finished 1491905729.9612 end 1491905729.9679 总时间为59.70270013809204
It takes 18 seconds for a single process to insert 10,000 records, which is less time-consuming than 10 processes inserting 10,000 records.
Inserting 100,000 records in a single process takes 187.40066790581, which is relatively slow. three minutes. . .
However, when I fork 1,000 processes to insert 100,000 records, an error may occur in about 36 seconds if successful. Mysqli_connection returns false. Is the number of connections limited?
Fork 10,000 child processes and insert 1 million data. At this time, there will be many connection errors. In the end, it took 360 seconds, and 945,300 records were inserted into the data table, with a success rate of 94.53%. So check the relevant configuration information of the database
mysql> show global status like '%connect%'; +-----------------------------------------------+---------------------+ | Variable_name | Value | +-----------------------------------------------+---------------------+ | Aborted_connects | 0 | | Connection_errors_accept | 0 | | Connection_errors_internal | 0 | | Connection_errors_max_connections | 628 | | Connection_errors_peer_address | 0 | | Connection_errors_select | 0 | | Connection_errors_tcpwrap | 0 | | Connections | 16519 | | Locked_connects | 0 | | Max_used_connections | 501 | | Max_used_connections_time | 2017-04-12 15:19:54 | | Performance_schema_session_connect_attrs_lost | 0 | | Ssl_client_connects | 0 | | Ssl_connect_renegotiates | 0 | | Ssl_finished_connects | 0 | | Threads_connected | 4 | +-----------------------------------------------+---------------------+ mysql> show global variables like '%connect%'; +-----------------------------------------------+--------------------+ | Variable_name | Value | +-----------------------------------------------+--------------------+ | character_set_connection | utf8mb4 | | collation_connection | utf8mb4_general_ci | | connect_timeout | 10 | | disconnect_on_expired_password | ON | | init_connect | | | max_connect_errors | 100 | | max_connections | 500 | | max_user_connections | 0 | | performance_schema_session_connect_attrs_size | 512 | +-----------------------------------------------+--------------------+ 修改 myqsql 配置文件，/etc/my.cnf 把max_connections 改为10000，然后重启mysql 实际MySQL服务器允许的最大连接数16384； 结果然并卵，虚拟机好像挂了了。
When the concurrency is large, the problem lies in connecting to mysql.
You can try to solve this problem through a connection pool.
In the virtual machine centos7, single core, 1G memory
/**
* 模拟并发请求，10万次写入数据库
* 拆分为10个进程，每个进程处理一万条插入
*/
$total = 10000;
$num = 10;
$per = $total/$num;
$sql = '';
$child = '';
echo 'start '.microtime(true).PHP_EOL;
for($i = 1; $i<= $num; $i++)
{
$pid = pcntl_fork();
if($pid == -1) {
die('fork error');
}
if($pid > 0) {
//$id = pcntl_wait($status,WNOHANG);
$child[] = $pid;
} else if ($pid == 0) {
$link = mysqli_connect('localhost','root','root','yii2advanced');
$start = ($i-1)*$per + 1;
$end = $start + $per;
for($j = $start; $j< $end; $j++){
$time = microtime(true);
$sql = 'insert pcntl_test (rank,time) values ('.$j.','.$time.')';
mysqli_query($link,$sql);
}
mysqli_close($link);
$id = getmypid();
echo 'child '.$id.' finished '.microtime(true).PHP_EOL;
exit(0);
}
}
while(count($child)){
foreach($child as $k => $pid) {
$res = pcntl_waitpid($pid, $status, WNOHANG);
if ( -1 == $res || $res > 0) {
unset($child[$k]);
}
}
}
echo 'end '.microtime(true).PHP_EOL;
When $total=10000, $num = 10; the execution result is as follows:
start 1491903371.5548 child 19860 finished 1491903417.2113 child 19857 finished 1491903417.6909 child 19864 finished 1491903417.7793 child 19855 finished 1491903417.8695 child 19859 finished 1491903417.9162 child 19861 finished 1491903418.0089 child 19856 finished 1491903418.0532 child 19863 finished 1491903418.0842 child 19862 finished 1491903418.1474 child 19858 finished 1491903418.4341 end 1491903418.4424 总时间为46.88759994506836秒
When $total= 10000, $num = 100, the execution result is as follows:
start 1491904334.1735 child 20085 finished 1491904337.0712 child 20086 finished 1491904337.144 …… child 20262 finished 1491904341.5602 child 20264 finished 1491904341.5803 end 1491904341.5869 总时间为7.413399934768677
When $total=10000,$num = 1000, the execution result is as follows:
start 1491904562.0166 child 20282 finished 1491904562.1191 child 20277 finished 1491904562.1268 child 20279 finished 1491904562.1352 ... child 21586 finished 1491904576.6954 child 21582 finished 1491904576.7024 child 21584 finished 1491904576.7226 end 1491904576.7297 总时间为14.71310019493103，相比100个子进程，耗时更长了。进程切换太多，影响了了效率应该是原因之一。
When $total=100000,$num =100, 100,000 records, 100 processes inserting
start 1491905670.2652 child 21647 finished 1491905725.4382 child 21651 finished 1491905725.4595 child 21642 finished 1491905725.5402 .... child 21810 finished 1491905729.7709 child 21812 finished 1491905729.8498 child 21811 finished 1491905729.9612 end 1491905729.9679 总时间为59.70270013809204
It takes 18 seconds for a single process to insert 10,000 records, which is less time-consuming than 10 processes inserting 10,000 records.
Inserting 100,000 records in a single process takes 187.40066790581, which is relatively slow. three minutes. . .
However, when I fork 1,000 processes to insert 100,000 records, an error may occur in about 36 seconds if successful. Mysqli_connection returns false. Is the number of connections limited?
Fork 10,000 sub-processes and insert 1 million data. At this time, there will be many connection errors. In the end, it took 360 seconds, and 945,300 records were inserted into the data table, with a success rate of 94.53%. So check the relevant configuration information of the database
mysql> show global status like '%connect%'; +-----------------------------------------------+---------------------+ | Variable_name | Value | +-----------------------------------------------+---------------------+ | Aborted_connects | 0 | | Connection_errors_accept | 0 | | Connection_errors_internal | 0 | | Connection_errors_max_connections | 628 | | Connection_errors_peer_address | 0 | | Connection_errors_select | 0 | | Connection_errors_tcpwrap | 0 | | Connections | 16519 | | Locked_connects | 0 | | Max_used_connections | 501 | | Max_used_connections_time | 2017-04-12 15:19:54 | | Performance_schema_session_connect_attrs_lost | 0 | | Ssl_client_connects | 0 | | Ssl_connect_renegotiates | 0 | | Ssl_finished_connects | 0 | | Threads_connected | 4 | +-----------------------------------------------+---------------------+ mysql> show global variables like '%connect%'; +-----------------------------------------------+--------------------+ | Variable_name | Value | +-----------------------------------------------+--------------------+ | character_set_connection | utf8mb4 | | collation_connection | utf8mb4_general_ci | | connect_timeout | 10 | | disconnect_on_expired_password | ON | | init_connect | | | max_connect_errors | 100 | | max_connections | 500 | | max_user_connections | 0 | | performance_schema_session_connect_attrs_size | 512 | +-----------------------------------------------+--------------------+ 修改 myqsql 配置文件，/etc/my.cnf 把max_connections 改为10000，然后重启mysql 实际MySQL服务器允许的最大连接数16384； 结果然并卵，虚拟机好像挂了了。
When the concurrency is large, the problem lies in connecting to mysql.
You can try to solve this problem through a connection pool.
The above is the detailed content of PHP multi-process insert data. For more information, please follow other related articles on the PHP Chinese website!