zabbix and many values.

[zeds]

I am the one who uses zabbix both as a monitoring tool and performance data gathering tool. Some of useful data is CPU load and io data.
See in the end of post you can see how I currently do monitoring of CPU data. I wrote a script which gets data, script which gets data for agent and then templates in zabbix server.
It worked quite OK.

But few days earlier I received SUN T3 dual procesoors server - it has 255 virtual CPUs(threads). mpstat output attached in the end. And I cannot force myself to make a new template with 255x15 items in it . There should be a better ways how to deal with such data. I thought that I could automate it by writing PHP script, but could not find database structure for 1.8.4 database.
Is it possible to write custom agent and feed a multiple data items at once ?
May be you have ideas how to gather performance data by not making any item by hand ?
The same applies to iodata to - when i have a few volumes attached to a server, it is ok to enter them one by one. But when I have do create 4 items for 64 volumes - it is becoming quite an annoing job.


This is how I do things now:
* * * * * /opt/zabbix/bin/mpstat_stat.bash 2>&1 >/dev/null

/etc/zabbix.conf
UserParameter=solaris.mcpu[*],/opt/zabbix/bin/grep_mpstat.pl $1 $2

This scripts gets mpstat data
Script may seem wird , but locking and copying is meant to protect us from
"item is not supported" when zabbix agent reads file in a moment it is being written to.

#!/bin/bash
# Script dumps solaris CPU data via mpstat to a file
# works on Solaris
NUMCPU=`/usr/sbin/psrinfo|wc -l|awk '{ print  $1 }'`;export NUMCPU
LOCK=/var/tmp/mpstat.lck
/usr/bin/mpstat 1 5|/usr/bin/tail -$NUMCPU  >/var/tmp/mpstat.txt.1
chown zabbix:zabbix /var/tmp/mpstat.txt.1
/bin/touch $LOCK
/bin/touch /var/tmp/mpstat.txt
chown zabbix:zabbix /var/tmp/mpstat.txt
cat /var/tmp/mpstat.txt.1 >/var/tmp/mpstat.txt 
/bin/rm -f $LOCK

and

This scripts read from previosly saved file. It provide better speed and efficiency.

#!/usr/bin/perl
#script greps /tmp/mpstat.txt file for changes
#it is designed to get load on individual CPU and cores
#for zabbix 
# script takes two parameters  core number and stat type

use warnings;
#use strict;
sub print_usage;

my $statfile='/var/tmp/mpstat.txt';
my $lockfile='/var/tmp/mpstat.lck';

#if DBGFLAG file exists , then diagnostic debug output is written to $dbgfile
$DBGFLAG='/etc/zabbix/debug_grep_mpstat';
my $dbgfile='/var/log/zabbix/grep_mpstat_debug.log';

if($#ARGV!=1) 
 {
 print_usage();
 exit(-1);
 }

$DBG=0;
if( -f $DBGFLAG ) { $DBG=1 };

my $core=shift;
my $stattype=shift;
my $stat_idx;


if( ! ( $core =~ /^[0-9]+$/ ))
 {
 print "invalid core number!\n";
 print_usage();
 exit(-2);
 }

my @stat_types=qw/minf mjf xcal  intr ithr  csw icsw migr smtx  srw syscl  usr sys  wt idl/;
my $bmatch=0;
my $stat_len=@stat_types;
my $i;
for($i=0;$i<$stat_len;$i++)
 {
 if($stat_types[$i] eq $stattype) 
   {
   $bmatch=1;
   $stat_idx=$i;
   }
 }

if($bmatch==0)
 {
 print "invalid stat type!\n";
 print_usage();
 exit(-3);
 }

my $nothing;
while( -f "$lockfile" )
 {
 $nothing++;
 }

$result="";
open FH,"<",$statfile or die "cannot open input file $statfile\n";
if( $DBG ) { open LOG,">>",$dbgfile or die "cannot open debug file $dbgfile"; };
while($line=readline(FH))
 {
 if( $DBG ) 
   {
   print LOG  "READ: $line\n";
   }
 $_=$line;
 s/^\s+//;
 s/\s+$//;
 $line=$_;
 
 my @sp=split(/\s+/,$line);
 if( $DBG )  
   {
   print LOG "SP array:";
   $i=0;
   foreach(@sp) 
     {
     print LOG "$i:$_  ";
     $i++;
     }
   print LOG "\n";
   };
 if( $DBG )  { print LOG "core=$core sp[0]=$sp[0]\n"};
 if($sp[0] !~ /^[0-9]+$/) {
   next;
  }; 
 
 if($core==$sp[0])
  {
  $result=$sp[$stat_idx+1];
  if( $DBG ) 
    {
    print LOG  "MATCH: $line\n";
    print LOG  "RESULT: $result\n";
    }
  }
 }
if ($DBG ) 
 { 
 close(DBG);
 }
close FH;
print "$result\n";

#subroutines
sub print_usage() {
print <<EOF;
 script greps /tmp/mpstat.txt file for required stat data
 usage:
  grep_mpstat.pl corenum [minf|mjf|xcal|intr|ithr|csw|icsw|migr|smtx|srw|syscl|usr|sys|wt|idl]
  
  Debugging:
   if file  $DBGFLAG exists debugging information is written in logfile $dbgfile
EOF
}

Attached mpstat output of dual T3. It quite interesting also - it shows solaris resource amanagement in action - CPU 128 -to 255 are assigned to a separate processor set , and solaris zone are allowed to use CPUs in this set only.