[SOLVED] perl script runs fine from command line but fails when invoked by nagios

Hi Folks,

this is the command line:
[root@monitor libexec]#./check_acc_bk.pl localhost 2
97 Files backed up in the last 2 days

this is the same command invoked by nagios:
[1349278228] EXTERNAL COMMAND: SCHEDULE_FORCED_SVC_CHECK;localhost;Check Account Backups;1349278227
[1349278238] SERVICE ALERT: localhost;Check Account Backups;CRITICAL;HARD;4;No files backed up in the previous 2 days


in nagios commands.cfg:

# 'check_acc_bk.pl' command definition
define command{
        command_name    check_acc_bk
        command_line    $USER1$/check_acc_bk.pl $HOSTADDRESS$ $ARG1$
        }

then in local.cfg:

define service{
        use                             local-service         ; Name of service template to use
        host_name                       localhost
        service_description             Check Account Backups
        active_checks_enabled           1
        check_interval                  15
        check_command                   check_acc_bk
        notifications_enabled           0
        }

and in check_acc_bk.pl:

#!/usr/bin/perl
use strict;
use File::Copy;
$|=1;
my $hostname = $ARGV[0] || die "Usage: (hostname) (days previous)\n";
my $delay = $ARGV[1] || '2' ; 
my $status="3";
my $msg="warning failed to run properly \n";
my $retval="0";
$delay = $delay -1;
#my $newfiles = changedfiles($delay);
&changedfiles($delay);
#print "The number of files changed in $delay is $retval\n";
chomp($retval);
#print STDERR"***$retval****in $delay days ******";
        if ($retval ne '0') {
                $status=0;
                $msg="$retval Files backed up in the last $delay days \n";
        } else {
                $delay = $delay +1;
                &changedfiles($delay);
                chomp($retval);
                #print STDERR"***$retval****in $delay days ******";
                #print "The number of files changed in $delay is $newfiles1 \n";
                if ($retval ne '0') {
                        $status=1;
                        $msg="$retval files backed up in the last $delay days \n";
                } else {
                        $status=2;
                        $msg="No files backed up in the  previous $delay days \n";
                }
        }


#=================
# Status Processing
#=================
# 3 UNKNOWN
# 2 CRITICAL
# 1 WARNING
# 0 OK
print $msg;
exit ($status);

#print  "======= subs  =================\n";
sub changedfiles($delay)
{
#       print"delay is $delay\n";
        my $com  = "sudo /usr/bin/find /root/accounting/backup/ -type f -mtime -$delay -print  | wc -l";
        my @lines = `$com`;
        my $num;
        foreach $num (@lines)
        {
                #print "the num = $num\n"; #put in for debugging
                if ($num > 0) {
                $retval = $num;
                }
        }

        return $retval;
}

oh and nagios has sudo rights on /usr/bin/find with no password.


I have another almost identical perl script that checks database file backups and it works fine.

Please help me Obiwan you are my last chance

Much easier to read if you use [code][/code] tags.

For debugging purposes can you output the value of $PATH at the beginning of the script? Not that I can see it should make a difference.

What distro are you running this on?

I don't quite understand this subroutine:

sub changedfiles($delay)
{
#       print"delay is $delay\n";
        my $com  = "sudo /usr/bin/find /root/accounting/backup/ -type f -mtime -$delay -print  | wc -l";
        my @lines = `$com`;
        my $num;
        foreach $num (@lines)
        {
                #print "the num = $num\n"; #put in for debugging
                if ($num > 0) {
                $retval = $num;
                }
        }

        return $retval;
}

Why are you allowing for more than one line to be returned?

cool thank you for the formatting tip - I was wondering how that was achieved - I'm new to forum participation but have been working with open source since 1994 (was my first Linux install). In those days we used Kermit via a dialup modem (9600 baud) to reach the internet and it was mostly text docs.

I program mostly in Java - Perl is something I use every occasionally (once a year maybe) so please excuse my ignorance. I am only trying to capture one value from the $com command - any suggestions will be appreciated.

The subroutine is to count the number of new and changed files in a directory.

(Centos 5.x) Linux monitor.jagmobile.co.za 2.6.18-308.11.1.el5PAE #1 SMP Tue Jul 10 09:29:33 EDT 2012 i686 i686 i386 GNU/Linux

Nagios Core 3.3.1

perl-5.8.8-38.el5

I believe the root of the problem is in the system call:

 my $com  = "sudo /usr/bin/find /root/accounting/backup/ -type f -mtime -$delay -print  | wc -l";

however the following is more Nagios friendly

my $com  = system("sudo /usr/bin/find /root/accounting/backup/ -type f -mtime -$delay -print  | wc -l");

but this causes the subroutine to return 0;
Ideas/examples on how to capture the system calls output to a returnable variable would be great.

FYI: I just found out that Perl at the command line is not the same as Nagios embedded Perl

What are the versions?
Put this in your script as a quick and dirty check

print `perl -v`

Since you need to capture the output of the wc command, stick with backticks. Also I think system() requires a list not a string, and may not directly support piping.

But what I would suggest is some more error checking ...

sub changedfiles($delay)
{
#       print"delay is $delay\n";
        my $com  = "sudo /usr/bin/find /root/accounting/backup/ -type f -mtime -$delay -print  | wc -l";
        my @lines = `$com`;
        my $rc = $?;
        if ($rc != 0) { die "Non-zero return code $rc from command"; }
        if ($#lines != 0) { die "Unexpected number of lines ".($#lines+1)." from command"; }

        $retval = $lines[0];
        return $retval;
}

Another thought: is Nagios running in a bash shell? Or at least some shell that supports pipes (|)? If not, it might be necessary to explicitly call bash -c "$com" or something like that.

It was a permissions issue - solved by:
1) moving the test to a remote machine using nrpe.
2) enabling sudo in nrpe.cfg.
3) configuring sudo to let nagios run perl scripting in libexec without a password.
4) and changing the monitoring machines config to accommodate the remote call.

WOOOOHOOOOO

OK, so none of my suggestions then
Glad you fixed it!