首页
学习
活动
专区
圈层
工具
发布
社区首页 >问答首页 >如何等待AnyEvent::HTTP中http_request的完成?

如何等待AnyEvent::HTTP中http_request的完成?
EN

Stack Overflow用户
提问于 2017-04-12 20:05:23
回答 1查看 322关注 0票数 1

程序从数据库中获取作业并执行AnyEvent::HTTP::http_request,在等待所有请求完成后,如何正确退出程序?

代码语言:javascript
复制
#!/usr/bin/perl
use strict;
use AnyEvent;
use AnyEvent::DBI;
use AnyEvent::HTTP;


my @queue;
my $added = 0;
my $finished = 0;
my $uid = int(rand(900000)) + 100000;
my ($run, $jobs, $log, $quit);

my $db = AnyEvent::DBI->new('DBI:mysql:dbname=;host=');

my $cv = AnyEvent->condvar;

$run = AnyEvent->timer(
    after => 5,
    interval => 0.1,
    cb => sub {
        if ($#queue != -1 && $added - $finished < 300) {
            my $job = shift @queue;
            my $r; $r = http_request(
                GET => $job->{url},
                sub {
                    undef $r;
                    my ($body, $header) = @_;
                    ...
                    $finished++;
                }
            );
            $added++;
        }
    }
);

$jobs = AnyEvent->timer(
    after => 0.1,
    interval => 5,
    cb => sub {
        if ($#queue < 1000) {
            $db->exec(q{UPDATE `jobs` SET `lock` = ? WHERE `lock` = 0 ORDER BY `time` ASC LIMIT 1000}, $uid, sub {
                $db->exec(q{SELECT * FROM `jobs` WHERE `lock` = ?}, $uid, sub {
                    my ($db, $rows, $rv) = @_;
                    push @queue, @$rows;
                    $db->exec(q{UPDATE `jobs` SET `lock` = 1 WHERE `lock` = ?}, $uid, sub { });
                });
            });
        }
    }
);

$log = AnyEvent->timer(
    after => 5,
    interval => 3,
    cb => sub {
        printf "Queue: %-6d Added: %-6d During: %-6d Total: %-6d\n", $#queue, $added, $added-$finished, $finished;
    }
);

$quit = AnyEvent->timer(
    after => 1,
    interval => 10,
    cb => sub {
         if (-f 'stop') {
            print "Exit\n";
            $cv->send;
         }
    }
);

my $result = $cv->recv;

也许您知道排队作业并执行它们的最佳方法,然后在AnyEvent + AnyEvent::HTTP中显示模板,我使用AnyEvent计时器,哪个更好、更快?

新版本:

代码语言:javascript
复制
#!/usr/bin/perl
use strict;
use AnyEvent;
use AnyEvent::DBI;
use AnyEvent::HTTP;

my $pid = int(rand(900000)) + 100000;
my $run = 1;
my $capacity = 300;
my $added = 0;
my $finished = 0;

my $cv = AnyEvent->condvar;

my $dbh = AnyEvent::DBI->new('DBI:mysql:dbname=;host=');

my $log = AnyEvent->timer(
    after => 1,
    interval => 3,
    cb => sub {
        printf "Added: %-6d Finished: %-6d Active: %-6d\n", $added, $finished, $AnyEvent::HTTP::ACTIVE if $finished;
        if ($run == 0 && $AnyEvent::HTTP::ACTIVE == 0) {
            $cv->send;
        }
    }
);

while (! -f 'stop') {
        my $done = AnyEvent->condvar;
        $dbh->exec(q{UPDATE `jobs` SET `lock` = ? WHERE `lock` = 0 ORDER BY `time` ASC LIMIT ?}, $pid, $capacity, sub {
            $dbh->exec(q{SELECT * FROM `jobs` WHERE `lock` = ?}, $pid, sub {
                my ($dbh, $rows, $rv) = @_;
                $dbh->exec(q{UPDATE `jobs` SET `lock` = 1 WHERE `lock` = ?}, $pid, sub {});
                $done->send($rows);
            });
        });
        my $jobs = $done->recv;

        my $done = AnyEvent->condvar;
        foreach my $job (@$jobs) {
            my $content;
            my $r; $r = http_request(
                GET =>  $job->{url},
                sub {
                    undef $r;
                    my ($body, $header) = @_;
                    ...
                    $dbh->exec(q{UPDATE `jobs` SET `lock` = 0, `time` = CURRENT_TIMESTAMP WHERE `id` = ?}, $job->{id}, sub {});
                    $done->send if $AnyEvent::HTTP::ACTIVE < $capacity;
                    $finished++;
                }
            );
            $added++;
        }
        $done->recv;
}
$run = 0;
$cv->recv;
EN

回答 1

Stack Overflow用户

回答已采纳

发布于 2017-04-13 20:46:52

你基本上需要一个三态制:

  1. 运行(接受新作业)。
  2. 退出(不接受新作业;等待现有作业完成)。
  3. 退出(不运行作业)。

您的两个版本都失败了,因为它们只有两个状态:

  1. 运行(接受新作业)
  2. 退出(作业可能仍在运行!)

这里有一个解决方案:

代码语言:javascript
复制
#!/usr/bin/perl
use strict;
use warnings;

use AE             qw( );
use AnyEvent       qw( );
use AnyEvent::DBI  qw( );
use AnyEvent::HTTP qw( );
use Scalar::Util   qw( weaken );
use Sys::Hostname  qw( hostname );

my $capacity = 300;

my $exit_request = AE::cv();
my $done         = AE::cv();
my $exiting      = 0;
my $grabbing     = 0;
my $added        = 0;
my $finished     = 0;

my $uid = join('.', hostname(), $$, int(rand(65536)));

my $db = AnyEvent::DBI->new('DBI:mysql:dbname=;host=');

sub worker {
    my ($job, $cb) = @_;
    http_request(
        GET =>  $job->{url},
        sub {
            my ($body, $header) = @_;
            ...
            $cb->();
        },
    );
}

sub manager {
    my $active = $added - $finshed;
    if ($exiting) {
        $done->send() if !$active;
        return;
    }

    my $avail_slots = $capacity - $active;
    return if !$avail_slots;

    return if $grabbing;

    $grabbing = 1;
    $db->exec(qq{UPDATE `jobs` SET `lock` = ? WHERE `status` = 'queued' AND `lock` = 0 ORDER BY `time` ASC LIMIT $avail_slots}, $uid, sub {
        $db->exec(q{SELECT * FROM `jobs` WHERE `status` = 'queued' AND `lock` = ?}, $uid, sub {
            my (undef, $jobs, undef) = @_;
            $db->exec(q{UPDATE `jobs` SET `status = 'wip' WHERE `lock` = ?}, $uid, sub {
                $grabbing = 0;
                for my $job (@$jobs) {
                    ++$added;
                    worker($job, sub {
                        ++$finished;
                        $db->exec(q{UPDATE `jobs` SET `status = 'done', `lock` = 0, `time` = CURRENT_TIMESTAMP WHERE `id` = ?}, $job->{id}, sub { });
                        manager();
                    });
                });
            });
        });
    });
};

my $db_poll_timer = AE::timer(5, 0.5, \&manager);

my $exit_check_timer = AE::timer(0, 2, sub {
    $exit_request->send() if -f 'stop';
});

my $log_timer = AE::timer(1, 3, sub {
    printf("Added: %-6d Finished: %-6d Active: %-6d\n",
        $added, $finished, $added-$finished);
});

$exit_request->recv();
print("Exiting...\n");
undef $exit_check_timer;
$exiting = 1;

$done->recv();
print("Finished.\n");
undef $db_poll_timer;
undef $log_timer;

功能:

  • 删除@queue,以避免从其他任务中窃取工作。
  • 我将员工代码与员工管理代码分开。
  • 我删除了工作人员是http_request的任何依赖项。
  • 通过基于计时器的轮询来支持长期运行的工作人员(比如第一个版本,但不像第二个版本)。
  • 使用$grabbing修正了抢占作业代码(在第一个版本中存在,但在第二个版本中不存在)中的竞争条件。

我超越了范围:

  • 使用了声发射的快捷方式。
  • 我用了一个更可靠的锁值。它现在是一个字符串,它甚至标识持有锁的机器和进程。
  • 我向作业表(status)中添加了一个字段,指示作业的状态,而不是重用lock字段。

今后的工作:

  • 使用AnyEvent::Filesys::Notify代替计时器进行文件检查?
票数 1
EN
页面原文内容由Stack Overflow提供。腾讯云小微IT领域专用引擎提供翻译支持
原文链接:

https://stackoverflow.com/questions/43378697

复制
相关文章

相似问题

领券
问题归档专栏文章快讯文章归档关键词归档开发者手册归档开发者手册 Section 归档