Perl系统调用引发奇怪的错误

Question

I just wrote a perl script that is restarting a list of services on a linux server. It's intended to run as a cron job. when I execute the script though, I keep getting this error;

root@www:~/scripts# ./ws_restart.pl
  * Stopping web server apache2                                   [ OK ]
sh: Syntax error: "(" unexpected
  * Stopping MySQL database server mysqld                         [ OK ]
sh: Syntax error: "(" unexpected

The call that is being used to do this is;

system("/etc/init.d/apache2 stop");
system("/etc/init.d/mysql stop");

I can paste the entire script code if needed, but I figured that this is the source of the problem and just need to know how to stop it.

Any ideas?

Here's the entire script;

#!/usr/bin/perl

use strict;
use warnings;

use Data::Dumper;

my $old_pids = {};
my $post_stop_ids = {};
my @services = qw/apache2 mysql solr/;
my $app_dir  = '/home/grip/apps/eventfinder';


# collect existing pids then kill services
foreach my $service (@services) {
  # gather up existing pids
  $old_pids->{$service} = [ get_pids_by_process($service) ];

  # issue stop command to each service
  set_service_state($service, 'stop');

  # attempt to regather same ids
  $post_stop_ids->{$service} = [ get_pids_by_process($service) ];

  # kill any rogue ids left over
  kill_rogue_procs($post_stop_ids->{$service});

  # give each kill time to finish
  sleep(5);
}


# attempt to restart killed services
foreach my $service (@services) {
  # issue start command to each service
  set_service_state($service, 'start');

  # Let's give each service enough time to crawl outta bed.
  # I know how much I hate waking up
  sleep(5);
}

# wait for it!...wait for it! :P
# Pad an extra 5 seconds to give solr enough time to come up before we reindex
sleep(5);

# start the reindexing process of solr
system("cd $app_dir ; RAILS_ENV=production rake reindex_active");

# call it a day...phew!
exit 0;

sub kill_rogue_procs {
  my @ids = shift;
  # check if we still have any rogue processes that failed to die
  # if so, kill them now.
  if(scalar @ids) {
    foreach my $pid (@ids) {
      system("kill $pid");
    }
  }
}


sub set_service_state {
  my ($proc, $state) = @_;

  if($proc eq 'apache2') {
    system("/etc/init.d/apache2 $state");
  } elsif($proc eq 'mysql') {
    system("/etc/init.d/mysql $state");
  } elsif($proc eq 'solr') {
    system("cd $app_dir ; RAILS_ENV=production rake sunspot:solr:$state");
  }
}


sub get_pids_by_process {
  my $proc = shift;

  my @proc_ids = ();

  open(PSAE, "/bin/ps -ae | grep $proc |") || die("Couldn't run command");

  while(<PSAE>) {
    push @proc_ids, $_ =~ /(\d{1,5})/;
  }

  close PSAE;

  return @proc_ids;
}

Answer 1

Actually, I'd be more suspicious of what's in @ids in kill_rogue_procs. It's the result of a ps followed by a grep, so might have bogus values if ps doesn't return any results or if the pid isn't 5 digits long.

Answer 2

This is wrong:

sub kill_rogue_procs {
  my @ids = shift;
  # check if we still have any rogue processes that failed to die
  # if so, kill them now.
  if(scalar @ids) {

From what you're passing to this sub, @ids will always contain a single array reference, so (scalar @ids) will always be true. It also means you end up passing something like the following to sh :

kill ARRAY(0x91b0768)

You want something like (if the arrayref is empty, there's nothing to loop over anyway):

my $ids = shift;
...
for my $pid (@$ids) {
  kill SIGTERM => $pid;

Or instead of the loop:

kill SIGTERM => @$ids;

Also, there is no need to call system to kill a process.

To this, I'd add the last line, so you don't grep the grep process itself:

sub get_pids_by_process {
  my $proc = shift;
  $proc =~ s/^(.)/[$1]/;  

Answer 3

As sh is raising the errors, I'm pretty sure one of the parameters to system is being expanded to something unexpected. I'd print all parameters just prior to passing them to system for a quick debug.