This file is indexed.

/usr/bin/check_mk_agent is in check-mk-agent 1.2.6p12-1.

This file is owned by root:root, with mode 0o755.

The actual contents of the file can be viewed below.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
#!/bin/bash
# +------------------------------------------------------------------+
# |             ____ _               _        __  __ _  __           |
# |            / ___| |__   ___  ___| | __   |  \/  | |/ /           |
# |           | |   | '_ \ / _ \/ __| |/ /   | |\/| | ' /            |
# |           | |___| | | |  __/ (__|   <    | |  | | . \            |
# |            \____|_| |_|\___|\___|_|\_\___|_|  |_|_|\_\           |
# |                                                                  |
# | Copyright Mathias Kettner 2014             mk@mathias-kettner.de |
# +------------------------------------------------------------------+
#
# This file is part of Check_MK.
# The official homepage is at http://mathias-kettner.de/check_mk.
#
# check_mk 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 in version 2.  check_mk is  distributed
# in the hope that it will be useful, but WITHOUT ANY WARRANTY;  with-
# out even the implied warranty of  MERCHANTABILITY  or  FITNESS FOR A
# PARTICULAR PURPOSE. See the  GNU General Public License for more de-
# ails.  You should have  received  a copy of the  GNU  General Public
# License along with GNU Make; see the file  COPYING.  If  not,  write
# to the Free Software Foundation, Inc., 51 Franklin St,  Fifth Floor,
# Boston, MA 02110-1301 USA.

# Remove locale settings to eliminate localized outputs where possible
export LC_ALL=C
unset LANG

export MK_LIBDIR="/usr/lib/check_mk_agent"
export MK_CONFDIR="/etc/check_mk"
export MK_VARDIR="/var/lib/check_mk_agent"

# Provide information about the remote host. That helps when data
# is being sent only once to each remote host.
if [ "$REMOTE_HOST" ] ; then
    export REMOTE=$REMOTE_HOST
elif [ "$SSH_CLIENT" ] ; then
    export REMOTE=${SSH_CLIENT%% *}
fi

# Make sure, locally installed binaries are found
PATH=$PATH:/usr/local/bin

# All executables in PLUGINSDIR will simply be executed and their
# ouput appended to the output of the agent. Plugins define their own
# sections and must output headers with '<<<' and '>>>'
PLUGINSDIR=$MK_LIBDIR/plugins

# All executables in LOCALDIR will by executabled and their
# output inserted into the section <<<local>>>. Please
# refer to online documentation for details about local checks.
LOCALDIR=$MK_LIBDIR/local

# All files in SPOOLDIR will simply appended to the agent
# output if they are not outdated (see below)
SPOOLDIR=$MK_VARDIR/spool

# close standard input (for security reasons) and stderr
if [ "$1" = -d ]
then
    set -xv
else
    exec </dev/null 2>/dev/null
fi

# Runs a command asynchronous by use of a cache file
function run_cached () {
    local section=
    if [ "$1" = -s ] ; then local section="echo '<<<$2>>>' ; " ; shift ; fi
    local NAME=$1
    local MAXAGE=$2
    shift 2
    local CMDLINE="$section$@"

    if [ ! -d $MK_VARDIR/cache ]; then mkdir -p $MK_VARDIR/cache ; fi
    CACHEFILE="$MK_VARDIR/cache/$NAME.cache"

    # Check if the creation of the cache takes suspiciously long and return
    # nothing if the age (access time) of $CACHEFILE.new is twice the MAXAGE
    local NOW=$(date +%s)
    if [ -e "$CACHEFILE.new" ] ; then
        local CF_ATIME=$(stat -c %X "$CACHEFILE.new")
        if [ $((NOW - CF_ATIME)) -ge $((MAXAGE * 2)) ] ; then
            # Kill the process still accessing that file in case
            # it is still running. This avoids overlapping processes!
            fuser -k -9 "$CACHEFILE.new" >/dev/null 2>&1
            rm -f "$CACHEFILE.new"
            return
        fi
    fi

    # Check if cache file exists and is recent enough
    if [ -s "$CACHEFILE" ] ; then
        local MTIME=$(stat -c %Y "$CACHEFILE")
        if [ $((NOW - MTIME)) -le $MAXAGE ] ; then local USE_CACHEFILE=1 ; fi
        # Output the file in any case, even if it is
        # outdated. The new file will not yet be available
        cat "$CACHEFILE"
    fi

    # Cache file outdated and new job not yet running? Start it
    if [ -z "$USE_CACHEFILE" -a ! -e "$CACHEFILE.new" ] ; then
        # When the command fails, the output is throws away ignored
        echo "set -o noclobber ; exec > \"$CACHEFILE.new\" || exit 1 ; $CMDLINE && mv \"$CACHEFILE.new\" \"$CACHEFILE\" || rm -f \"$CACHEFILE\" \"$CACHEFILE.new\"" | nohup bash >/dev/null 2>&1 &
    fi
}

# Make run_cached available for subshells (plugins, local checks, etc.)
export -f run_cached

echo '<<<check_mk>>>'
echo Version: 1.2.6p12
echo AgentOS: linux
echo AgentDirectory: $MK_CONFDIR
echo DataDirectory: $MK_VARDIR
echo SpoolDirectory: $SPOOLDIR
echo PluginsDirectory: $PLUGINSDIR
echo LocalDirectory: $LOCALDIR

# If we are called via xinetd, try to find only_from configuration
if [ -n "$REMOTE_HOST" ]
then
    echo -n 'OnlyFrom: '
    echo $(sed -n '/^service[[:space:]]*check_mk/,/}/s/^[[:space:]]*only_from[[:space:]]*=[[:space:]]*\(.*\)/\1/p' /etc/xinetd.d/* | head -n1)
fi

# Print out Partitions / Filesystems. (-P gives non-wrapped POSIXed output)
# Heads up: NFS-mounts are generally supressed to avoid agent hangs.
# If hard NFS mounts are configured or you have too large nfs retry/timeout
# settings, accessing those mounts from the agent would leave you with
# thousands of agent processes and, ultimately, a dead monitored system.
# These should generally be monitored on the NFS server, not on the clients.

echo '<<<df>>>'
# The exclusion list is getting a bit of a problem. -l should hide any remote FS but seems
# to be all but working.
excludefs="-x smbfs -x cifs -x iso9660 -x udf -x nfsv4 -x nfs -x mvfs -x zfs"
df -PTlk $excludefs | sed 1d

# df inodes information
echo '<<<df>>>'
echo '[df_inodes_start]'
df -PTli $excludefs | sed 1d
echo '[df_inodes_end]'

# Filesystem usage for ZFS
if type zfs > /dev/null 2>&1 ; then
    echo '<<<zfsget>>>'
    zfs get -Hp name,quota,used,avail,mountpoint,type -t filesystem,volume || \
       zfs get -Hp name,quota,used,avail,mountpoint,type
    echo '[df]'
    df -PTlk -t zfs | sed 1d
fi

# Check NFS mounts by accessing them with stat -f (System
# call statfs()). If this lasts more then 2 seconds we
# consider it as hanging. We need waitmax.
if type waitmax >/dev/null
then
    STAT_VERSION=$(stat --version | head -1 | cut -d" " -f4)
    STAT_BROKE="5.3.0"

    echo '<<<nfsmounts>>>'
    sed -n '/ nfs4\? /s/[^ ]* \([^ ]*\) .*/\1/p' < /proc/mounts |
        sed 's/\\040/ /g' |
        while read MP
	do
	 if [ $STAT_VERSION != $STAT_BROKE ]; then
	    waitmax -s 9 2 stat -f -c "$MP ok %b %f %a %s" "$MP" || \
		echo "$MP hanging 0 0 0 0"
	 else
	    waitmax -s 9 2 stat -f -c "$MP ok %b %f %a %s" "$MP" && \
	    printf '\n'|| echo "$MP hanging 0 0 0 0"
	 fi
	done

    echo '<<<cifsmounts>>>'
    sed -n '/ cifs\? /s/[^ ]* \([^ ]*\) .*/\1/p' < /proc/mounts |
        sed 's/\\040/ /g' |
        while read MP
        do
         if [ $STAT_VERSION != $STAT_BROKE ]; then
            waitmax -s 9 2 stat -f -c "$MP ok %b %f %a %s" "$MP" || \
                echo "$MP hanging 0 0 0 0"
         else
            waitmax -s 9 2 stat -f -c "$MP ok %b %f %a %s" "$MP" && \
            printf '\n'|| echo "$MP hanging 0 0 0 0"
         fi
        done
fi

# Check mount options. Filesystems may switch to 'ro' in case
# of a read error.
echo '<<<mounts>>>'
grep ^/dev < /proc/mounts

# processes including username, without kernel processes
echo '<<<ps>>>'
ps ax -o user,vsz,rss,cputime,pid,command --columns 10000 | sed -e 1d -e 's/ *\([^ ]*\) *\([^ ]*\) *\([^ ]*\) *\([^ ]*\) *\([^ ]*\) */(\1,\2,\3,\4,\5) /'

# Memory usage
echo '<<<mem>>>'
egrep -v '^Swap:|^Mem:|total:' < /proc/meminfo

# Load and number of processes
echo '<<<cpu>>>'
echo "$(cat /proc/loadavg) $(grep -E '^CPU|^processor' < /proc/cpuinfo | wc -l)"

# Uptime
echo '<<<uptime>>>'
cat /proc/uptime


# New variant: Information about speed and state in one section
echo '<<<lnx_if:sep(58)>>>'
sed 1,2d /proc/net/dev
if type ethtool > /dev/null
then
    for eth in $(sed -e 1,2d < /proc/net/dev | cut -d':' -f1 | sort)
    do
        echo "[$eth]"
        ethtool $eth | egrep '(Speed|Duplex|Link detected|Auto-negotiation):'
        echo -en "\tAddress: " ; cat /sys/class/net/$eth/address ; echo
    done
fi


# Current state of bonding interfaces
if [ -e /proc/net/bonding ] ; then
    echo '<<<lnx_bonding:sep(58)>>>'
    pushd /proc/net/bonding > /dev/null ; head -v -n 1000 * ; popd
fi

# Same for Open vSwitch bonding
if type ovs-appctl > /dev/null ; then
    echo '<<<ovs_bonding:sep(58)>>>'
    for bond in $(ovs-appctl bond/list | sed -e 1d | cut -f2) ; do
        echo "[$bond]"
        ovs-appctl bond/show $bond
    done
fi


# Number of TCP connections in the various states
echo '<<<tcp_conn_stats>>>'
# waitmax 10 netstat -nt | awk ' /^tcp/ { c[$6]++; } END { for (x in c) { print x, c[x]; } }'
# New implementation: netstat is very slow for large TCP tables
cat /proc/net/tcp /proc/net/tcp6 2>/dev/null | awk ' /:/ { c[$4]++; } END { for (x in c) { print x, c[x]; } }'

# Linux Multipathing
if type multipath >/dev/null ; then
    if [ -f /etc/multipath.conf ] ; then
        echo '<<<multipath>>>'
        multipath -l
    fi
fi

# Performancecounter Platten
echo '<<<diskstat>>>'
date +%s
egrep ' (x?[shv]d[a-z]*|cciss/c[0-9]+d[0-9]+|emcpower[a-z]+|dm-[0-9]+|VxVM.*|mmcblk.*) ' < /proc/diskstats
if type dmsetup >/dev/null ; then
    echo '[dmsetup_info]'
    dmsetup info -c --noheadings --separator ' ' -o name,devno,vg_name,lv_name
fi
if [ -d /dev/vx/dsk ] ; then
    echo '[vx_dsk]'
    stat -c "%t %T %n" /dev/vx/dsk/*/*
fi


# Performancecounter Kernel
echo '<<<kernel>>>'
date +%s
cat /proc/vmstat /proc/stat

# Hardware sensors via IPMI (need ipmitool)
if type ipmitool > /dev/null
then
    run_cached -s ipmi 300 "ipmitool sensor list | grep -v 'command failed' | sed -e 's/ *| */|/g' -e 's/ /_/g' -e 's/_*"'$'"//' -e 's/|/ /g' | egrep -v '^[^ ]+ na ' | grep -v ' discrete '"
fi


# IPMI data via ipmi-sensors (of freeipmi). Please make sure, that if you
# have installed freeipmi that IPMI is really support by your hardware.
if type ipmi-sensors >/dev/null
then
    echo '<<<ipmi_sensors>>>'
    # Newer ipmi-sensors version have new output format; Legacy format can be used
    if ipmi-sensors --help | grep -q legacy-output; then
        IPMI_FORMAT="--legacy-output"
    else
        IPMI_FORMAT=""
    fi
    # At least with ipmi-sensoirs 0.7.16 this group is Power_Unit instead of "Power Unit"
    run_cached -s ipmi_sensors 300 "for class in Temperature Power_Unit Fan
    do
        ipmi-sensors $IPMI_FORMAT --sdr-cache-directory /var/cache -g "$class" | sed -e 's/ /_/g' -e 's/:_\?/ /g' -e 's@ \([^(]*\)_(\([^)]*\))@ \2_\1@'
        # In case of a timeout immediately leave loop.
        if [ $? = 255 ] ; then break ; fi
    done"
fi

# RAID status of Linux software RAID
echo '<<<md>>>'
cat /proc/mdstat

# RAID status of Linux RAID via device mapper
if type dmraid >/dev/null && DMSTATUS=$(dmraid -r)
then
    echo '<<<dmraid>>>'

    # Output name and status
    dmraid -s | grep -e ^name -e ^status

    # Output disk names of the RAID disks
    DISKS=$(echo "$DMSTATUS" | cut -f1 -d\:)

    for disk in $DISKS ; do
        device=$(cat /sys/block/$(basename $disk)/device/model )
        status=$(echo "$DMSTATUS" | grep ^${disk})
        echo "$status Model: $device"
    done
fi

# RAID status of LSI controllers via cfggen
if type cfggen > /dev/null ; then
   echo '<<<lsi>>>'
   cfggen 0 DISPLAY | egrep '(Target ID|State|Volume ID|Status of volume)[[:space:]]*:' | sed -e 's/ *//g' -e 's/:/ /'
fi

# RAID status of LSI MegaRAID controller via MegaCli. You can download that tool from:
# http://www.lsi.com/downloads/Public/MegaRAID%20Common%20Files/8.02.16_MegaCLI.zip
if type MegaCli >/dev/null ; then
    MegaCli_bin="MegaCli"
elif type MegaCli64 >/dev/null ; then
    MegaCli_bin="MegaCli64"
elif type megacli >/dev/null ; then
    MegaCli_bin="megacli"
else
    MegaCli_bin="unknown"
fi

if [ "$MegaCli_bin" != "unknown" ]; then
    echo '<<<megaraid_pdisks>>>'
    for part in $($MegaCli_bin -EncInfo -aALL -NoLog < /dev/null \
        | sed -rn 's/:/ /g; s/[[:space:]]+/ /g; s/^ //; s/ $//; s/Number of enclosures on adapter ([0-9]+).*/adapter \1/g; /^(Enclosure|Device ID|adapter) [0-9]+$/ p'); do
        [ $part = adapter ] && echo ""
        [ $part = 'Enclosure' ] && echo -ne "\ndev2enc"
        echo -n " $part"
    done
    echo
    $MegaCli_bin -PDList -aALL -NoLog < /dev/null | egrep 'Enclosure|Raw Size|Slot Number|Device Id|Firmware state|Inquiry|Adapter'
    echo '<<<megaraid_ldisks>>>'
    $MegaCli_bin -LDInfo -Lall -aALL -NoLog < /dev/null | egrep 'Size|State|Number|Adapter|Virtual'
    echo '<<<megaraid_bbu>>>'
    $MegaCli_bin -AdpBbuCmd -GetBbuStatus -aALL -NoLog < /dev/null | grep -v Exit
fi

# RAID status of 3WARE disk controller (by Radoslaw Bak)
if type tw_cli > /dev/null ; then
    for C in $(tw_cli show | awk 'NR < 4 { next } { print $1 }'); do
        echo '<<<3ware_info>>>'
        tw_cli /$C show all | egrep 'Model =|Firmware|Serial'
        echo '<<<3ware_disks>>>'
        tw_cli /$C show drivestatus | egrep 'p[0-9]' | sed "s/^/$C\//"
        echo '<<<3ware_units>>>'
        tw_cli /$C show unitstatus | egrep 'u[0-9]' | sed "s/^/$C\//"
    done
fi

# RAID controllers from areca (Taiwan)
# cli64 can be found at ftp://ftp.areca.com.tw/RaidCards/AP_Drivers/Linux/CLI/
if type cli64 >/dev/null ; then
    run_cached -s arc_raid_status 300 "cli64 rsf info | tail -n +3 | head -n -2"
fi

# VirtualBox Guests. Section must always been output. Otherwise the
# check would not be executed in case no guest additions are installed.
# And that is something the check wants to detect
echo '<<<vbox_guest>>>'
if type VBoxControl >/dev/null 2>&1 ; then
    VBoxControl -nologo guestproperty enumerate | cut -d, -f1,2
    [ ${PIPESTATUS[0]} = 0 ] || echo "ERROR"
fi

# OpenVPN Clients. Currently we assume that the configuration # is in
# /etc/openvpn. We might find a safer way to find the configuration later.
if [ -e /etc/openvpn/openvpn-status.log ] ; then
    echo '<<<openvpn_clients:sep(44)>>>'
    sed -n -e '/CLIENT LIST/,/ROUTING TABLE/p' < /etc/openvpn/openvpn-status.log  | sed -e 1,3d -e '$d'
fi

# Time synchronization with NTP
if type ntpq > /dev/null 2>&1 ; then
   # remove heading, make first column space separated
   run_cached -s ntp 30 "waitmax 5 ntpq -np | sed -e 1,2d -e 's/^\(.\)/\1 /' -e 's/^ /%/' || true"
fi

# Time synchronization with Chrony
if type chronyc > /dev/null 2>&1 ; then
    # Force successful exit code. Otherwise section will be missing if daemon not running
    run_cached -s chrony 30 "waitmax 5 chronyc tracking || true"
fi

if type nvidia-settings >/dev/null && [ -S /tmp/.X11-unix/X0 ]
then
    echo '<<<nvidia>>>'
    for var in GPUErrors GPUCoreTemp
    do
        DISPLAY=:0 waitmax 2 nvidia-settings -t -q $var | sed "s/^/$var: /"
    done
fi

if [ -e /proc/drbd ]; then
  echo '<<<drbd>>>'
  cat /proc/drbd
fi

# Status of CUPS printer queues
if type lpstat > /dev/null 2>&1; then
    if pgrep -x cups > /dev/null 2>&1; then
        # first define a function to check cups
        function cups_queues () {
            CPRINTCONF=/etc/cups/printers.conf
            if [ -r "$CPRINTCONF" ] ; then
                LOCAL_PRINTERS=$(grep -E "<(Default)?Printer .*>" $CPRINTCONF | awk '{print $2}' | sed -e 's/>//')
                lpstat -p | while read LINE
                do
                    PRINTER=$(echo $LINE | awk '{print $2}')
                    if echo "$LOCAL_PRINTERS" | grep -q "$PRINTER"; then
                        echo $LINE
                    fi
                done
                echo '---'
                lpstat -o | while read LINE
                do
                    PRINTER=${LINE%%-*}
                    if echo $LOCAL_PRINTERS | grep -q $PRINTER; then
                        echo $LINE
                    fi
                done
            else
                lpstat -p
                echo '---'
                lpstat -o | sort
                fi
        }
        # Make cups_queues available for subshell
        export -f cups_queues
        # Use cups_queues function with run_cached and cache time of 5 mins
        run_cached -s cups_queues 300 "cups_queues"
    fi
fi

# Heartbeat monitoring
# Different handling for heartbeat clusters with and without CRM
# for the resource state
if [ -S /var/run/heartbeat/crm/cib_ro -o -S /var/run/crm/cib_ro ] || pgrep crmd > /dev/null 2>&1; then
  echo '<<<heartbeat_crm>>>'
  crm_mon -1 -r | grep -v ^$ | sed 's/^ //; /^\sResource Group:/,$ s/^\s//; s/^\s/_/g'
fi
if type cl_status > /dev/null 2>&1; then
  echo '<<<heartbeat_rscstatus>>>'
  cl_status rscstatus

  echo '<<<heartbeat_nodes>>>'
  for NODE in $(cl_status listnodes); do
    if [ $NODE != $(echo $HOSTNAME | tr 'A-Z' 'a-z') ]; then
      STATUS=$(cl_status nodestatus $NODE)
      echo -n "$NODE $STATUS"
      for LINK in $(cl_status listhblinks $NODE 2>/dev/null); do
        echo -n " $LINK $(cl_status hblinkstatus $NODE $LINK)"
      done
      echo
    fi
  done
fi

# Postfix mailqueue monitoring
#
# Only handle mailq when postfix user is present. The mailq command is also
# available when postfix is not installed. But it produces different outputs
# which are not handled by the check at the moment. So try to filter out the
# systems not using postfix by searching for the postfix user.a
#
# Cannot take the whole outout. This could produce several MB of agent output
# on blocking queues.
# Only handle the last 6 lines (includes the summary line at the bottom and
# the last message in the queue. The last message is not used at the moment
# but it could be used to get the timestamp of the last message.
if type postconf >/dev/null ; then
    echo '<<<postfix_mailq>>>'
    postfix_queue_dir=$(postconf -h queue_directory)
    postfix_count=$(find $postfix_queue_dir/deferred -type f | wc -l)
    postfix_size=$(du -ks $postfix_queue_dir/deferred | awk '{print $1 }')
    if [ $postfix_count -gt 0 ]
    then
       echo -- $postfix_size Kbytes in $postfix_count Requests.
    else
       echo Mail queue is empty
    fi
elif [ -x /usr/sbin/ssmtp ] ; then
    echo '<<<postfix_mailq>>>'
    mailq 2>&1 | sed 's/^[^:]*: \(.*\)/\1/' | tail -n 6
fi

#Check status of qmail mailqueue
if type qmail-qstat >/dev/null
then
   echo "<<<qmail_stats>>>"
   qmail-qstat
fi

# Check status of OMD sites
if type omd >/dev/null
then
    run_cached -s omd_status 60 "omd status --bare --auto || true"
fi


# Welcome the ZFS check on Linux
# We do not endorse running ZFS on linux if your vendor doesnt support it ;)
# check zpool status
if type zpool >/dev/null; then
   echo "<<<zpool_status>>>"
   zpool status -x
fi


# Fileinfo-Check: put patterns for files into /etc/check_mk/fileinfo.cfg
if [ -r "$MK_CONFDIR/fileinfo.cfg" ] ; then
    echo '<<<fileinfo:sep(124)>>>'
    date +%s
    for line in $(cat "$MK_CONFDIR/fileinfo.cfg")
    do
        stat -c "%n|%s|%Y" $line 2>/dev/null

        if [ $? -ne 0 ]; then
            echo "$line|missing|$(date +%s)"
        fi
    done
fi

# Get stats about OMD monitoring cores running on this machine.
# Since cd is a shell builtin the check does not affect the performance
# on non-OMD machines.
if cd /omd/sites
then
    echo '<<<livestatus_status:sep(59)>>>'
    for site in *
    do
        if [ -S "/omd/sites/$site/tmp/run/live" ] ; then
            echo "[$site]"
            echo -e "GET status" | waitmax 3 /omd/sites/$site/bin/unixcat /omd/sites/$site/tmp/run/live
        fi
    done
fi

# Get statistics about monitored jobs. Below the job directory there
# is a sub directory per user that ran a job. That directory must be
# owned by the user so that a symlink or hardlink attack for reading
# arbitrary files can be avoided.
if pushd $MK_VARDIR/job >/dev/null; then
    echo '<<<job>>>'
    for username in *
    do
        if [ -d "$username" ] && cd "$username" ; then
            if [ $EUID -eq 0 ]; then
                su "$username" -c "head -n -0 -v *"
            else
                head -n -0 -v *
            fi
            cd ..
        fi
    done
    popd > /dev/null
fi

# Gather thermal information provided e.g. by acpi
# At the moment only supporting thermal sensors
if ls /sys/class/thermal/thermal_zone* >/dev/null 2>&1; then
    echo '<<<lnx_thermal>>>'
    for F in /sys/class/thermal/thermal_zone*; do
        echo -n "${F##*/} "
        if [ ! -e $F/mode ] ; then echo -n "- " ; fi
        cat $F/{mode,type,temp,trip_point_*} | tr \\n " "
        echo
    done
fi

# Libelle Business Shadow
if type trd >/dev/null; then
   echo "<<<libelle_business_shadow:sep(58)>>>"
   trd -s
fi

# MK's Remote Plugin Executor
if [ -e "$MK_CONFDIR/mrpe.cfg" ]
then
    echo '<<<mrpe>>>'
    grep -Ev '^[[:space:]]*($|#)' "$MK_CONFDIR/mrpe.cfg" | \
    while read descr cmdline
    do
        PLUGIN=${cmdline%% *}
        OUTPUT=$(eval "$cmdline")
        echo -n "(${PLUGIN##*/}) $descr $? $OUTPUT" | tr \\n \\1
        echo
    done
fi


# Local checks
echo '<<<local>>>'
if cd $LOCALDIR ; then
    for skript in $(ls) ; do
        if [ -f "$skript" -a -x "$skript" ] ; then
            ./$skript
        fi
    done
    # Call some plugins only every X'th minute
    for skript in [1-9]*/* ; do
        if [ -x "$skript" ] ; then
            run_cached local_${skript//\//\\} ${skript%/*} "$skript"
        fi
    done
fi

# Plugins
if cd $PLUGINSDIR ; then
    for skript in $(ls) ; do
        if [ -f "$skript" -a -x "$skript" ] ; then
            ./$skript
        fi
    done
    # Call some plugins only every Xth minute
    for skript in [1-9]*/* ; do
        if [ -x "$skript" ] ; then
            run_cached plugins_${skript//\//\\} ${skript%/*} "$skript"
        fi
    done
fi

# Agent output snippets created by cronjobs, etc.
if [ -d "$SPOOLDIR" ]
then
    pushd "$SPOOLDIR" > /dev/null
    now=$(date +%s)

    for file in *
    do
        # output every file in this directory. If the file is prefixed
        # with a number, then that number is the maximum age of the
        # file in seconds. If the file is older than that, it is ignored.
        maxage=""
        part="$file"

        # Each away all digits from the front of the filename and
        # collect them in the variable maxage.
        while [ "${part/#[0-9]/}" != "$part" ]
        do
            maxage=$maxage${part:0:1}
            part=${part:1}
        done

        # If there is at least one digit, than we honor that.
        if [ "$maxage" ] ; then
            mtime=$(stat -c %Y "$file")
            if [ $((now - mtime)) -gt $maxage ] ; then
                continue
            fi
        fi

        # Output the file
        cat "$file"
    done
    popd > /dev/null
fi