Munin contributed plugins

#!/usr/bin/perl -w

=HEADER

        -== Munin plugin for VMware ESXi/vSphere monitoring ==-

 Copyright (c) 2012 - Stefan Seidel <munin@stefanseidel.info>

    This program is free software: you can redistribute it and/or modify

    it under the terms of the GNU General Public License as published by

    the Free Software Foundation, either version 3 of the License, or

    (at your option) any later version.

    This program is distributed in the hope that it will be useful,

    but WITHOUT ANY WARRANTY; without even the implied warranty of

    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License

    along with this program.  If not, see <http://www.gnu.org/licenses/>.

 This plugin uses the vSphere SDK for Perl available at

 http://www.vmware.com/support/developer/viperltoolkit/

 or included in the vSphere CLI available at

 http://www.vmware.com/support/developer/vcli/

 The use of the SDK is subject to the terms and condition

 of VMware, Inc. to which you must agree upon installation.

=cut

=USAGE

       -== Usage ==-

Put this file in /usr/share/munin/plugins, `chmod +x` it and

`ln -s` it to /etc/munin/plugins/esx_<hostname of server to monitor>

Add a file "esx_" /etc/munin/plugin-conf.d with content like this

(omit the "# " at the beginning of each line)

---- snip ----

[esx_*]

timeout 60

env.user <username on ESX server or vCenter>

env.password <password of user on ESX server or vCenter>

---- snip ----

Then you need to add this host to your munin.conf on the munin server

(often this is the same as your munin node, i.e. this host) and restart

munin-node, and wait for the data to populate.

       -== Query all hosts of a vCenter ==-

An alternate way of using this plugin is to link it to

/etc/munin/plugins/vcenter_<hostname_of_vcenter>

In this mode, the vCenter server specified in the filename is queried

for a list of hosts it manages, and graphs are created for all these

hosts. The option "flatview" is implied in this mode, since one munin

plugin can only be assigned to one host. The option "vCenter" is ignored

since the vCenter server is given through the file name. You can, however

still use the option "flatview" to override the "host_name" under which

the results are reported. Make sure to read the section below about

this option!

       -== Graphs don't render ==-

Munin 1.4 has a bug with complex multigraphs like this, see

http://munin-monitoring.org/ticket/1224 for details and a fix if

your graphs don't render!

       -== Option flatview ==-

There is an option to render all VMs and Host Systems in a flat

structure, i.e. not rendering VMs as sub-items of their host.

This is useful if you frequently move VMs between hosts and want to

keep the VM graphs running. To activate this option, add

---- snip ----

env.flatview top_level_entry

---- snip ----

to the entry in your config file in /etc/munin/plugin-conf.d (see above).

Be aware that this has some drawbacks:

 - you cannot have the same VM name in two hosts you monitor

   (the VM name is the unique identifier for the graphs)

 - you will only indirectly be able to see which VM is on which host

   (running VMs will appear in the CPU graphs of their hosts)

 - it's a flat structure, so it can become quite a long list

 - because of the way Munin works, all hosts will be queried serially,

   not in parallel as it would be the case without "flat view" - this

   MAY lead to timing problems if you have a large number of hosts or VMs

       -== Option vCenter ==-

If you wish to access the host system indirectly through a vCenter, just

specify this parameter:

---- snip ----

env.vCenter <address or hostname of the vCenter>

---- snip ----

This option can be used with or without the "flatview" option. Make sure your

password and username are valid on the vCenter. The plugin name will still have

to contain the hostname of the host you want to monitor - be aware that you have

to use the hostname exactly as it is registered in the vCenter, so IPs and

hostnames are NOT interchangeable.

=cut

=ACK

     -== Ackknowledgements ==-

I would like to thank VMware for their SDK and the good documentation.

Special thanks go to MEGABIT Informationstechnik GmbH (www.megabit.net)

who graciously sponsored the development of the "flat view" option

and the ability to access hosts via vCenter as well as the feature to

query all hosts on the vCenter.

=cut

use strict;

use sort 'stable'; # guarantee stability

no warnings; # don't want warnings in output

use VMware::VIRuntime; # need to install VIM SDK (vSphere CLI/SDK 4.1 or newer)

use VMware::VILib;

use VMware::VIExt;

use Data::Dumper;

use DateTime::Format::ISO8601; # may need to install "libdatetime-format-iso8601-perl" on Debian-based systems

use List::Util qw(sum max);

use List::MoreUtils qw(all);

use Munin::Plugin;

use Time::HiRes qw(time);

my $DEBUG = ${Munin::Plugin::DEBUG};

# Important: this is needed if you do not use a "proper" SSL certificate

# on your vSphere/vCenter/ESX(i) server (which is the default)

$ENV{PERL_LWP_SSL_VERIFY_HOSTNAME} = 0;

# for datetime parsing later on

my $iso8601 = DateTime::Format::ISO8601->new;

my @host_names = ();

my $host_name;

my $host_view;

my $dtsys;

my $perfMan;

my %perfCounter;

# IDs/UUIDs to human readable names

my $resolveNames;

if ($0 =~ /vcenter_(.+)$/) {

    $ENV{vCenter} = $1;

    $ENV{flatview} = $ENV{flatview} || $ENV{vCenter};

    my $vpid = open(FH, "-|");

    if ($vpid == 0) {

        Opts::set_option ('username', $ENV{user} || 'root');

        Opts::set_option ('password', $ENV{password} || '');

        Opts::set_option ('server', $ENV{vCenter});

        Util::connect();

        foreach (@{Vim::find_entity_views(view_type => 'HostSystem', properties => ['name'])}) {

            print $_->{name}, "\n";

        }

        Util::disconnect();

        exit 0;

    } else {

        while (<FH>) {

            push @host_names, trim($_);

            print "# found host ",trim($_)," on vCenter\n" if $DEBUG;

        }

        close FH;

    }

} else {

    # get hostname from filename and blurt it out immediately

    # so that when something goes wrong, at least the plugin

    # output is linked with the right host

    $0 =~ /esx_(.+)$/;

    push @host_names, $1;

}

my @returns = ();

foreach $host_name (@host_names) {

    # make sure we only print one host_name statement per plugin

    if ((@returns == 0) and (defined $ARGV[0]) and ($ARGV[0] eq "config")) {

        if ($ENV{flatview}) {

            print "host_name $ENV{flatview}\n";

            print "# for host $host_name\n" if $DEBUG;

        } else {

            print "host_name $host_name\n";

        }

    }

    local *FH;

    my $pid = open(FH, "-|");

    if ($pid == 0) {

        # CHILD

        # env.user and env.password need to be set in plugin-conf.d

        Opts::set_option ('username', $ENV{user} || 'root');

        Opts::set_option ('password', $ENV{password} || '');

        if ($ENV{vCenter}) {

            print "# vCenter: $ENV{vCenter} - host $host_name\n" if $DEBUG;

            Opts::add_options ( (vihost => { alias => "h", type => "=s", required => 0 }) );

            Opts::set_option ('vihost',$host_name);

            Opts::set_option ('server',$ENV{vCenter});

        } else {

            Opts::set_option ('server',$host_name);

        }

        # plugin needs Munin 1.4 or later

        need_multigraph();

        my $sstarttime = time();

        # connect to vSphere host

        Util::connect();

        # central object host_view holds all relevant items (VMs, network, etc.)

        $host_view = VIExt::get_host_view(1, ['summary', 'network', 'datastore', 'vm', 'runtime', 'configManager.networkSystem', 'configManager.dateTimeSystem']);

        Opts::assert_usage(defined($host_view), "Invalid host.");

        my $serviceInst = Vim::get_view (mo_ref => ManagedObjectReference->new(type => 'ServiceInstance', value => 'ServiceInstance'));

        # Performance Manager for getting the actual values

        $perfMan = Vim::get_view (mo_ref => $serviceInst->content->perfManager);

        Opts::assert_usage(defined($perfMan), "No PerformanceManager.");

        # may be needed later

        #my $netsys = Vim::get_view(mo_ref => ManagedObjectReference->new(type => 'HostNetworkSystem', value => 'networkSystem'));

        #Opts::assert_usage(defined($netsys), "No NetworkSystem.");

        # used for getting the current vSphere server time and then

        # defining the (now - 5minutes) interval

        $dtsys = Vim::get_view(mo_ref => $host_view->{'configManager.dateTimeSystem'});

        Opts::assert_usage(defined($dtsys), "No DateTimeSystem.");

        print "# time to connect and get objects: ", time() - $sstarttime, "\n" if $DEBUG;

        # enumerate all performance counters by their IDs

        %perfCounter = map { $_->key => $_ } @{$perfMan->perfCounter};

        # holds all performance data

        my @all_perf_data = ();

        # store VM ids for iteration later on

        my @all_vms = ();

        $host_view->update_view_data();

        # retrieve performance counters for host

        push @all_perf_data, get_perf_data($host_view);

        # manually set UF name for host system

        $resolveNames->{vm}->{""} = "Host System";

        # only purpose of this loop is getting the UF network names

        # network ManagedObjects do not have performance counters

        for ($host_view->network) {

            for (@$_) {

                my $network = Vim::get_view (mo_ref => $_);

                $resolveNames->{net}->{$_->{value}} = $_->{value}." (".$network->summary->name.")";

            }

        }

        # purpose of this loop is getting the UF datastore names

        # and retrieving capacity and free/uncommitted space

        # datastore ManagedObjects do not have performance counters

        for ($host_view->datastore) {

            for (@$_) {

                my $datastore = Vim::get_view (mo_ref => $_);

                # update freeSpace values (doesn't work on free ESXi)

                eval { $datastore->RefreshDatastore(); };

                $datastore->update_view_data();

                my $uuid =$datastore->summary->url;

                $uuid =~ s!.+/!!;

                $resolveNames->{datastore}->{$uuid} = $datastore->name;

                push (@all_perf_data,

                            { rollup => "latest",

                               group => "datastore",

                                name => "capacity",

                               value => $datastore->summary->capacity,

                             counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Capacity", summary => "Maximum amount of storage space on this datastore")),

                                  vm => "",

                            instance => $uuid,

                                unit => "Bytes" });

                push (@all_perf_data,

                            { rollup => "latest",

                               group => "datastore",

                                name => "freeSpace",

                               value => $datastore->summary->freeSpace,

                             counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Free", summary => "Total amount of unused, available storage space on this datastore")),

                                  vm => "",

                            instance => $uuid,

                                unit => "Bytes" });

                push (@all_perf_data,

                            { rollup => "latest",

                               group => "datastore",

                                name => "uncommitted",

                               value => $datastore->summary->uncommitted,

                             counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Uncommitted", summary => "Total additional storage space, potentially used by all virtual machines on this datastore")),

                                  vm => "",

                            instance => $uuid,

                                unit => "Bytes" });

            }

        }

        # iterate over all vms

        for ($host_view->vm) {

            for (@$_) {

                my $vm = Vim::get_view (mo_ref => $_);

                $vm->update_view_data();

                # store VM id for later iteration

                my $vmId = $_->{value};

                push @all_vms, $vmId;

                # ID to VM name

                $resolveNames->{vm}->{$vmId} = "VM ".$vm->summary->config->name;

                $resolveNames->{vmuuid}->{$vmId} = $vm->summary->config->uuid;

                # fetch disk space usage per datastore

                for (@{$vm->storage->perDatastoreUsage}) {

                    my $uuid = Vim::get_view(mo_ref => $_->datastore)->summary->url;

                    $uuid =~ s!.+/!!;

                    push (@all_perf_data,

                                { rollup => "latest",

                                   group => "datastore",

                                    name => "committed",

                                   value => $_->committed,

                                 counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Comitted", summary => "Storage space, in bytes, on this datastore that is actually being used by the virtual machine.\n\nIt includes space actually occupied by disks, logs, snapshots, configuration files etc. Files of the virtual machine which are present on a different datastore (e.g. a virtual disk on another datastore) are not included here.\n\n")),

                                      vm => $vmId,

                                instance => $uuid,

                                    unit => "Bytes" });

                    push (@all_perf_data,

                                { rollup => "latest",

                                   group => "datastore",

                                    name => "uncommitted",

                                   value => $_->uncommitted,

                                 counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Uncomitted", summary => "Additional storage space, in bytes, potentially used by the virtual machine on this datastore.\n\nAdditional space may be needed for example when lazily allocated disks grow, or storage for swap is allocated when powering on the virtual machine.\n\nIf the virtual machine is running off delta disks (for example because a snapshot was taken), then only the potential growth of the currently used delta-disks is considered.\n\n")),

                                      vm => $vmId,

                                instance => $uuid,

                                    unit => "Bytes" });

                    push (@all_perf_data,

                                { rollup => "latest",

                                   group => "datastore",

                                    name => "unshared",

                                   value => $_->unshared,

                                 counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => "Unshared", summary => "Storage space, in bytes, occupied by the virtual machine on this datastore that is not shared with any other virtual machine.\n\n")),

                                      vm => $vmId,

                                instance => $uuid,

                                    unit => "Bytes" });

                }

                # retrieve performance counters for this VM

                push @all_perf_data, get_perf_data ($_);

            }

        }

        # keep track of how many sensors are in which state

        my %sensorCount = ( green => 0, red => 0, unknown => 0, yellow => 0 );

        # iterate over all sensor data

        my $index = 0;

        for (@{$host_view->runtime->healthSystemRuntime->systemHealthInfo->numericSensorInfo}) {

            # update counters

            $sensorCount{$_->healthState->key}++;

            # do not create entries for unmonitorable things like software components

            next unless ($_->baseUnits =~ /.+/);

            # create entry with sensor data

            push (@all_perf_data,

                        { rollup => "latest",

                           group => "sensors",

                            name => "sensor_".($index++),

                           value => $_->currentReading,

                         counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => $_->name, summary => "Sensor data for the ".$_->sensorType." sensor ".$_->name.". ".$_->healthState->summary." (".$_->healthState->label.")")),

                              vm => "",

                        instance => "",

                    unitModifier => $_->unitModifier,

                            unit => $_->baseUnits });

        }

        # we're finished querying the server, so we can disconnect now

        Util::disconnect();

        # create entries for the green/red/yellow/unknown counters

        for (keys %sensorCount) {

            push (@all_perf_data,

                        { rollup => "latest",

                           group => "sensors",

                            name => $_."_sensors",

                           value => $sensorCount{$_},

                         counter => PerfCounterInfo->new(nameInfo => ElementDescription->new(label => ucfirst($_), summary => "Count of sensors in the $_ state")),

                              vm => "",

                        instance => "",

                            unit => "Numbers" });

        }

        if ($DEBUG) {

            foreach (sort { $a->{group} cmp $b->{group} || $a->{instance} cmp $b->{instance} || $a->{name} cmp $b->{name} || $a->{rollup} cmp $b->{rollup} || $a->{vm} cmp $b->{vm} } @all_perf_data) {

                print "# $_->{vm}\t$_->{rollup}\t$_->{group}\t$_->{instance}\t$_->{name}\t$_->{value}\t$_->{unit}\n";

            }

        }

        # which graphs to draw

        my @all_graphs = ();

        # host system

        push @all_graphs, (

            {   selector => { group => qr/^cpu$/i, name => qr/^usagemhz$/i, instance => qr/^$/ },

                  config => { groupBy => "group", graphName => "host_cpu", graphTitle => "CPU usage per " }

            },

            {   selector => { group => qr/^disk$/i, name => qr/^(read|usage|write)$/i, instance => qr/.+/ },

                  config => { groupBy => "group", graphName => "host_disk_transfer", graphTitle => "Disk Transfer Rates per " }

            },

            {   selector => { group => qr/^disk$/i, name => qr/^.+Averaged$/i, instance => qr/.+/ },

                  config => { groupBy => "group", graphName => "host_disk_iops", graphTitle => "Disk I/O operations per " }

            },

            {   selector => { group => qr/^disk$/i, name => qr/^.+Latency$/i, instance => qr/.+/, vm => qr/^$/ },

                  config => { groupBy => "vm", graphName => "host_disk_latency", graphTitle => "Disk latency for " }

            },

            {   selector => { group => qr/^mem$/i, unit => qr/^KB$/i, rollup => qr/^none$/, vm => qr/^$/ },

                  config => { groupBy => "vm", graphName => "host_memory", graphTitle => "Memory usage for " }

            },

            {   selector => { group => qr/^datastore$/i, unit => qr/^Bytes$/i, vm => qr/^$/ },

                  config => { groupBy => "vm", graphName => "usage_datastore", graphTitle => "Disk space usage for ", graphArgs => "--lower-limit 10737418240 --logarithmic --alt-autoscale-min --units=si" }

            },

            {   selector => { group => qr/^net$/i, unit => qr/^KBps$/i, vm => qr/^$/ },

                  config => { groupBy => "vm", graphName => "host_traffic_net", graphTitle => "Network traffic for " }

            },

            {   selector => { group => qr/^net$/i, unit => qr/^Number$/i, vm => qr/^$/ },

                  config => { groupBy => "vm", graphName => "host_packets_net", graphTitle => "Network packets for " }

            },

            {   selector => { group => qr/^power$/i, name => qr/^power$/i },

                  config => { groupBy => "group", graphName => "power_usage", graphTitle => "Host System and VM " }

            },

            {   selector => { group => qr/^sys$/i, name => qr/^diskUsage$/i },

                  config => { groupBy => "name", graphName => "host_disk_usage", graphTitle => "Host System " }

            },

            {   selector => { group => qr/^sys$/i, name => qr/^uptime$/i },

                  config => { groupBy => "name", graphName => "uptimes", graphTitle => "Host System and VM ", graphArgs => "--lower-limit 1000 --logarithmic --alt-autoscale-min" }

            }

        );

        # graphs per VM

        foreach (@all_vms) {

            my $vmName = clean_fieldname($resolveNames->{vm}->{$_});

            push @all_graphs, (

                {   selector => { group => qr/^cpu$/i, name => qr/^usagemhz$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_cpu", graphTitle => "CPU usage for " }

                },

                {   selector => { group => qr/^mem$/i, unit => qr/^KB$/i, rollup => qr/^none$/, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_memory", graphTitle => "Memory usage for " }

                },

                {   selector => { group => qr/^datastore$/i, unit => qr/^Bytes$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_datastore", graphTitle => "Disk space usage for ", graphArgs => "--lower-limit 10485760 --logarithmic --alt-autoscale-min --units=si" }

                },

                {   selector => { group => qr/^virtualDisk$/i, unit => qr/^Millisecond$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_disklat", graphTitle => "Disk latency for " }

                },

                {   selector => { group => qr/^virtualDisk$/i, unit => qr/^Number$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_diskiops", graphTitle => "Disk I/O operations for " }

                },

                {   selector => { group => qr/^virtualDisk$/i, unit => qr/^KBps$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_disktrans", graphTitle => "Disk transfer rates for " }

                },

                {   selector => { group => qr/^net$/i, unit => qr/^KBps$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_traffic_net", graphTitle => "Network traffic for " }

                },

                {   selector => { group => qr/^net$/i, unit => qr/^Number$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_packets_net", graphTitle => "Network packets for " }

                },

                {   selector => { group => qr/^sys$/i, name => qr/^uptime$/i, vm => qr/^$_$/ },

                      config => { groupBy => "vm", graphName => "$vmName.vm_uptime", graphTitle => "VM uptime " }

                }

            );

        }

        # sensor graphs

        push @all_graphs, (

            {   selector => { group => qr/^sensors$/i },

                  config => { groupBy => "unit", graphName => "sensor_", graphTitle => "Sensors ", multiGraph => 1 }

            });

        print "# time to collect all data: ", time() - $sstarttime, "\n" if $DEBUG;

        # actual processing

        foreach (@all_graphs) {

            if ((defined $ARGV[0]) and ($ARGV[0] eq "config")) {

                munin_print("config", \@all_perf_data, $_);

                munin_print("values", \@all_perf_data, $_) if $ENV{MUNIN_CAP_DIRTYCONFIG}; # this doesn't seem to work even on Munin 1.4.6

            } else {

                munin_print("values", \@all_perf_data, $_);

            }

        }

        print "# time of the script: ", time() - $sstarttime, "\n" if $DEBUG;

        exit 0;

    } else {

        # PARENT

        push @returns, *FH;

    }

}

# gather and print the output of the forked processes

foreach my $fh (@returns) {

    while (<$fh>) {

        print $_;

    }

    close ($fh);

}

exit 0;

####################################################################

# calculate sum, max or avg from performance data values

sub process_value_array {

    my $arr = shift;

    my $pd = shift;

    my @vs = ();

    if ($pd->unitInfo->key eq "percent") {

        @vs = map { $_ / 100 } @$arr ;

    } else {

        @vs = @$arr;

    }

    return sum(@vs) if $pd->rollupType->val eq "summation";

    return max(@vs) if $pd->nameInfo->key =~ /max/i;

    return sum(@vs)/@$arr;

}

# query performance data for object

sub get_perf_data {

    my $entity = shift;

    my @ret = ();

    my $gathstart = time();

    # get the current server time

    my $curtime = $iso8601->parse_datetime($dtsys->QueryDateTime());

    # and subtract 5 minutes to get all values for the last period

    my $oldtime = $curtime->clone->add(minutes => -5);

    # actual query, intervalId is 20 because that's the default

    my $perfQ = $perfMan->QueryPerf(querySpec => PerfQuerySpec->new(entity => $entity, intervalId => 20, startTime => $oldtime));

    # loop over PerfEntityMetric

    foreach (defined $perfQ ? @$perfQ : ()) {

        my $vm = ($_->entity->type eq 'VirtualMachine')?$_->entity->value:"";

        # loop over PerfMetricIntSeries

        foreach (@{$_->{value}}) {

            my $perfDesc = $perfCounter{$_->id->counterId};

            next unless defined $perfDesc;

            push @ret, { rollup => $perfDesc->rollupType->val,

                          group => $perfDesc->groupInfo->key,

                           name => $perfDesc->nameInfo->key,

                          value => process_value_array(\@{$_->{value}}, $perfDesc),

                        counter => $perfDesc,

                             vm => $vm,

                       instance => $_->id->instance,

                           unit => $perfDesc->unitInfo->label };

        }

    }

    print "# time to gather info for $entity :", time() - $gathstart, "\n" if $DEBUG;

    return @ret;

}

# generate a munin-friendly and unique field name

sub gen_dp_name {

    my $fname = $_[0]->{name};

    $fname .= "v".$resolveNames->{vmuuid}->{$_[0]->{vm}} unless $_[1] eq "vm" or $_[0]->{vm} eq "";

    $fname .= "i$_[0]->{instance}" unless $_[1] eq "instance" or $_[0]->{instance} eq "";

    return clean_fieldname($fname);

}

# trim white spaces

sub trim {

    my $string = shift;

    $string =~ s/^\s+//;

    $string =~ s/\s+$//;

    return $string;

}

# print values and configs for graphs

sub munin_print {

    # action

    my $act = shift || "";

    # values

    my $arr = shift || ();

    # parameters

    my $par = shift || {};

    my $cfg = $par->{config};

    $par = $par->{selector};

    my $oldGroup = "_-_";

    my $factor;

    if ($ENV{flatview}) {

        $cfg->{graphName} = clean_fieldname("Host_".$host_view->name).".".$cfg->{graphName} unless $cfg->{graphName} =~ m/\./;

    }

    # find values according to criteria in $par and sort by grouping parameter

    #foreach (sort { $a->{$cfg->{groupBy}} cmp $b->{$cfg->{groupBy}} } grep { my $d = $_; all { (not exists $d->{$_}) || $d->{$_} =~ /$par->{$_}/ } keys %$par; } @$arr) {

     foreach (sort { $a->{$cfg->{groupBy}} cmp $b->{$cfg->{groupBy}} } grep { my $d = $_; all { (not exists $d->{$_}) || $d->{$_} =~ /$par->{$_}/ } keys %$par; } @$arr) {

        my $groupCrit = $cfg->{groupBy} || "";

        my $curGroup = $_->{$groupCrit} || "";

        if (!($curGroup eq $oldGroup)) {

            # we're in a new group, meaning a new graph starts

            $factor = 0;

            # clean up group name for multigraph name

            my $ccurGroup = $curGroup;

            $ccurGroup =~ s/ |\./_/g;

            print "multigraph ",$cfg->{graphName},(exists $cfg->{multiGraph}?$ccurGroup:""),"\n";

            if ("config" eq $act) {

                # want configuration

                print "graph_title ",$cfg->{graphTitle},$resolveNames->{$groupCrit}->{$curGroup} || $curGroup,"\n";

                #print "graph_order xxx yyy\n";

                my $unit = $_->{unit};

                my $base = 1000;

                # since the y-axis markers are going to be wrong with source units like

                # KB, MB, MHz etc., we define a correction factor via cdef later

                # this way, if 1024 MB is reported, the graph shows 1G and not 1k

                # (although 1k MB is technically also correct, but confusing)

                if ($unit =~ /^Bytes$/i) {

                    $base = 1024;

                } elsif ($unit =~ /^KBps$/i) {

                    $unit = "Bytes/s";

                    $factor = 1024;

                    $base = 1024;

                } elsif ($unit =~ /^KB$/i) {

                    $unit = "Bytes";

                    $factor = 1024;

                    $base = 1024;

                } elsif ($unit =~ /^MB$/i) {

                    $unit = "Bytes";

                    $factor = 1024*1024;

                    $base = 1024;

                } elsif ($unit =~ /^MHz$/i) {

                    $unit = "Hz";

                    $factor = 1000000;

                } elsif ($unit =~ /^Millisecond$/i) {

                    $unit = "Second";

                    $factor = 1/1000;

                }

                print "graph_vlabel $unit\n";

                print "graph_category $_->{group}\n";

                print "graph_args --base=$base --alt-autoscale-max ",(defined $cfg->{graphArgs})?$cfg->{graphArgs}:"","\n";

            }

        }

        $oldGroup = $curGroup;

        my $dpName = gen_dp_name($_, $groupCrit);

        if ("config" eq $act) {

            # want configuration

            # get instance and VM names and UF names, if applicable

            my $iName = $resolveNames->{$_->{group}}->{$_->{instance}} || (("" eq $_->{instance})?"":$_->{group}." ".$_->{instance});

            $iName = " $iName" if $iName;

            my $vmName = $resolveNames->{vm}->{$_->{vm}};

            $vmName = " $vmName" if $vmName;

            # all values are drawn as lines for now

            print "$dpName.draw LINE2\n";

            print "$dpName.label ",$_->{counter}->nameInfo->label,$iName,("vm" eq $groupCrit)?"":$vmName || "","\n";

            my $summary = $_->{counter}->nameInfo->summary;

            $summary =~ s!\n!\\n!g;

            print "$dpName.info ",$summary,$iName?", instance$iName ($_->{instance})":"",$vmName?",$vmName":"","\n";

            # declare CDEF if we want to apply a factor

            if ($factor > 1) {

                print "$dpName.cdef $dpName,$factor,*\n";

            } elsif ($factor <= 0) {

                if (defined $_->{unitModifier}) {

                    # sensor values have a unit modifier M attached to them so that REALVAL=VAL*10^M

                    # y,x,LOG,*,EXP is x^y, just in case this is not obvious to the reader

                    print "$dpName.cdef $dpName,",$_->{unitModifier},",10,LOG,*,EXP,*\n";

                }

            } elsif ($factor < 1) {

                print "$dpName.cdef $dpName,",1/$factor,",/\n";

            }

        } else {

            # just print value

            print "$dpName.value $_->{value}\n";

        }

    }

}