場景
在程序中遇到耗時較長的操作（如等待用戶響應，連接remote server），通常會想到創建 Child process 去處理。在Perl 中 使用fork()
## fork a child process
die "$@" unless defined( my $child_pid = fork());
if ($child_pid) { # If I have a child PID, then I must be the parent
push @children_pids, $child_pid;
print "children's PIDs: @children_pids\n";
} else { # I am the child
my $wait_time = int(rand(10));
sleep $wait_time;
my $localtime = localtime;
print "Child: Some child exited at $localtime\n";
exit 0; # Exit the child
}
那么 Child process 的回收問題怎么解決？如果 Parent process 任由它自生自滅，就會導致Zombie process。
回收方法
有兩種方式如下，無論哪種都涉及waitpid(),當然也可以使用wait()（不建議）
- Blocking wait
- Non blocking wait
在此提一下waitpid()下文code有使用
waitpid PID,FLAGS
PID: -1(stands for any child process) | pid ( > 0 )
FLAGS: 0 (for a blocking wait) | WHOHANG (return 0 or -1if no dead children)
i.e,. waitpid($pid, 0) ## in a blocking wait
waitpid(-1, WHOHANG) ## no blocking wait
Blocking wait
見如下code:
my @children_pids;
print "Parent: my pid $$\n";
for my $count (1..10){
die "$@" unless defined( my $child_pid = fork());
if ($child_pid) { # If I have a child PID, then I must be the parent
push @children_pids, $child_pid;
print "children's PIDs: @children_pids\n";
} else { # I am the child
my $wait_time = int(rand(10));
sleep $wait_time;
my $localtime = localtime;
print "Child: Some child exited at $localtime\n";
exit 0; # Exit the child
}
}
foreach my $child (@children_pids) {
print "Parent: Waiting on $child\n";
waitpid($child, 0); ## will not go to next step unless $child reaped
my $localtime = localtime;
print "Parent: Child $child was reaped - $localtime.\n";
}
Output:
[root@VTB93-PC1 ~]# perl /tmp/test_fork.pl
Parent: my pid 15117
children's PIDs: 15118
children's PIDs: 15118 15119
children's PIDs: 15118 15119 15120
children's PIDs: 15118 15119 15120 15121
children's PIDs: 15118 15119 15120 15121 15122
children's PIDs: 15118 15119 15120 15121 15122 15123
children's PIDs: 15118 15119 15120 15121 15122 15123 15124
children's PIDs: 15118 15119 15120 15121 15122 15123 15124 15125
children's PIDs: 15118 15119 15120 15121 15122 15123 15124 15125 15126
children's PIDs: 15118 15119 15120 15121 15122 15123 15124 15125 15126 15127
Parent: Waiting on 15118
Child: Some child exited at Thu Apr 21 13:28:48 2016
Parent: Child 15118 was reaped - Thu Apr 21 13:28:48 2016.
Parent: Waiting on 15119
Child: Some child exited at Thu Apr 21 13:28:48 2016
Child: Some child exited at Thu Apr 21 13:28:48 2016
Child: Some child exited at Thu Apr 21 13:28:50 2016
Child: Some child exited at Thu Apr 21 13:28:51 2016
Parent: Child 15119 was reaped - Thu Apr 21 13:28:51 2016.
Parent: Waiting on 15120
Child: Some child exited at Thu Apr 21 13:28:51 2016
Child: Some child exited at Thu Apr 21 13:28:52 2016
Child: Some child exited at Thu Apr 21 13:28:53 2016
Child: Some child exited at Thu Apr 21 13:28:54 2016
Parent: Child 15120 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15121
Parent: Child 15121 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15122
Parent: Child 15122 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15123
Child: Some child exited at Thu Apr 21 13:28:54 2016
Parent: Child 15123 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15124
Parent: Child 15124 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15125
Parent: Child 15125 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15126
Parent: Child 15126 was reaped - Thu Apr 21 13:28:54 2016.
Parent: Waiting on 15127
Parent: Child 15127 was reaped - Thu Apr 21 13:28:54 2016.
可以看出規律，回收的順序是按照｀fork()｀的順序, 可在實際中 child process 耗時有長有短,后fork()的process很有可能比較早的先結束，所以引入第二種Reap機制，使用$SIG{CHLD}
Non blocking wait
這里先提一下$SIG{CHLD}, 可接受的賦值如下：
$SIG{CHLD} = 'IGNORE'; ## Children reaped by system
$SIG{CHLD} = 'DEFAULT'; ## System defined
$SIG{CHLD} = &REAPER; ## do REAPER if SIGCHLD catched
use POSIX ":sys_wait_h";
$SIG{CHLD}=\&REAPER;
sub REAPER {
my $child;
while(( $child = waitpid(-1, &WNOHANG)) > 0){
my $localtime = localtime;
print "Parent: Child $child was reaped - $localtime.\n";
}
$SIG{CHLD}=\&REAPER;
}
my @children_pids;
print "Parent: my pid $$\n";
for my $count (1..10){
die "$@" unless defined( my $child_pid = fork());
if ($child_pid) { # If I have a child PID, then I must be the parent
push @children_pids, $child_pid;
print "children's PIDs: @children_pids\n";
} else { # I am the child
my $wait_time = int(rand(10));
sleep $wait_time;
my $localtime = localtime;
print "Child: Some child exited at $localtime\n";
exit 0; # Exit the child
}
}
## Keep parent alive to reap all children
while (1) {
sleep;
}
Outpuit:
Parent: my pid 15189
children's PIDs: 15194
children's PIDs: 15194 15195
children's PIDs: 15194 15195 15196
children's PIDs: 15194 15195 15196 15197
children's PIDs: 15194 15195 15196 15197 15198
children's PIDs: 15194 15195 15196 15197 15198 15199
children's PIDs: 15194 15195 15196 15197 15198 15199 15200
children's PIDs: 15194 15195 15196 15197 15198 15199 15200 15201
children's PIDs: 15194 15195 15196 15197 15198 15199 15200 15201 15202
children's PIDs: 15194 15195 15196 15197 15198 15199 15200 15201 15202 15203
Child: Some child exited at Thu Apr 21 13:42:25 2016
Parent: Child 15202 was reaped - Thu Apr 21 13:42:25 2016.
Child: Some child exited at Thu Apr 21 13:42:27 2016
Parent: Child 15197 was reaped - Thu Apr 21 13:42:27 2016.
Child: Some child exited at Thu Apr 21 13:42:29 2016
Parent: Child 15201 was reaped - Thu Apr 21 13:42:29 2016.
Child: Some child exited at Thu Apr 21 13:42:30 2016
Parent: Child 15194 was reaped - Thu Apr 21 13:42:30 2016.
Child: Some child exited at Thu Apr 21 13:42:31 2016
Parent: Child 15198 was reaped - Thu Apr 21 13:42:31 2016.
Child: Some child exited at Thu Apr 21 13:42:31 2016
Parent: Child 15200 was reaped - Thu Apr 21 13:42:31 2016.
Child: Some child exited at Thu Apr 21 13:42:31 2016
Parent: Child 15203 was reaped - Thu Apr 21 13:42:31 2016.
Child: Some child exited at Thu Apr 21 13:42:32 2016
Parent: Child 15199 was reaped - Thu Apr 21 13:42:32 2016.
Child: Some child exited at Thu Apr 21 13:42:33 2016
Parent: Child 15195 was reaped - Thu Apr 21 13:42:33 2016.
Child: Some child exited at Thu Apr 21 13:42:34 2016
Parent: Child 15196 was reaped - Thu Apr 21 13:42:34 2016.
可以看出但某個Child exit 后會發送SIGCHLD給Parent，順序依據exit先后。當然，這是Parent跟蹤Children（如獲取返回信息local $?，以上都返回 0 ）作出相應操作。如果無需如此，可直接$SIG{CHLD}='INGNORE'; 交給OS吧。
總結
- 使用 SIGCHLD REAPER 更符合現實場景，結合
waitpid()以 Non blocking wait 方式 - Parent 負責 Children 的生命周期，以免出現 Zombie process
- PID 1 init ## 所有Process 的 parent
