Module: check_mk Branch: master Commit: 4ceb309e259023ad5c329a032354162b27c40e72 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=4ceb309e259023… Author: Mathias Kettner &lt;mk(a)mathias-kettner.de&gt; Date: Thu Mar 17 09:48:02 2011 +0100 diskstat: rewritten check: now show different devices --- ChangeLog | 1 + checkman/diskstat | 2 +- checks/diskstat | 99 +++++++++++++++++++++++----------- pnp-templates/check_mk-diskstat.php | 48 ++++++++++++++--- web/plugins/perfometer/check_mk.py | 13 +++++ web/plugins/views/perfometer.py | 28 ++++++++++ 6 files changed, 151 insertions(+), 40 deletions(-) diff --git a/ChangeLog b/ChangeLog index d6de430..a368655 100644 --- a/ChangeLog +++ b/ChangeLog @@ -34,6 +34,7 @@ with index 1 * df: split PNP graphs for growth/trend into two graphs * omd_status: new check for checking status of OMD sites + * diskstat: rewritten check: now show different devices, r+w in one check 1.1.10: diff --git a/checkman/diskstat b/checkman/diskstat index 098bbb9..9ef5a5b 100644 --- a/checkman/diskstat +++ b/checkman/diskstat @@ -1,4 +1,4 @@ -title: Measure Disk IO +title: Monitor disk throughput on Linux agents: linux author: Mathias Kettner &lt;mk(a)mathias-kettner.de&gt; license: GPL diff --git a/checks/diskstat b/checks/diskstat index 8819cd2..e4bff2b 100644 --- a/checks/diskstat +++ b/checks/diskstat @@ -24,46 +24,81 @@ # to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, # Boston, MA 02110-1301 USA. +# <<<diskstat>>> +# 1300264105 +# 8 0 sda 691860 951191 13559915 491748 234686 197346 3359512 94944 0 56844 586312 +# 8 32 sdb 791860 91191 23589915 491748 234686 197346 3359512 94944 0 56844 586312 -def inventory_diskstat(checkname, info): - if len(info) > 1: # first line is timestamp - return [('read', "", "None"), - ('write', "", "None")] - -# Fields in /proc/diskstat -# At the beginning: major minor devicename -# -# Field 1 -- # of reads issued -# Field 2 -- # of reads merged, field 6 -- # of writes merged -# Field 3 -- # of sectors read -# Field 4 -- # of milliseconds spent reading -# Field 5 -- # of writes completed -# Field 7 -- # of sectors written -# Field 8 -- # of milliseconds spent writing -# Field 9 -- # of I/Os currently in progress -# Field 10 -- # of milliseconds spent doing I/Os -# Field 11 -- weighted # of milliseconds spent doing I/Os +# Fields in /proc/diskstats +# Index 0 -- major number +# Index 1 -- minor number +# Index 2 -- device name +# Index 3 -- # of reads issued +# Index 4 -- # of reads merged, field 6 -- # of writes merged +# Index 5 -- # of sectors read (a 512 Byte) +# Index 6 -- # of milliseconds spent reading +# Index 7 -- # of writes completed +# Index 8 -- # of sectors written (a 512 Byte) +# Index 9 -- # of milliseconds spent writing +# Index 10 -- # of I/Os currently in progress +# Index 11 -- # of milliseconds spent doing I/Os +# Index 12 -- weighted # of milliseconds spent doing I/Os # # -> Field 1 has index 3 +diskstat_default_levels = { +# "read" : (10, 20), # MB/sec +# "write" : (20, 40), # MB/sec +# "average" : 15, # min +} -# major minor device ... -def check_diskstat(item, params, info): - # sum up over all devices - if item == 'read': - index = 5 # sectors read - elif item == 'write': - index = 9 # sectors written - else: - return (3, "UNKNOWN - invalid item %s" % (item,)) +def inventory_diskstat(checkname, info): + inventory = [] + for line in info[1:]: + inventory.append( (line[2], "diskstat_default_levels" ) ) + return inventory +def check_diskstat(item, params, info): + average_range = params.get("average") this_time = int(info[0][0]) - this_val = sum([int(x[index]) for x in info[1:]]) + for line in info[1:]: + if line[2] == item: + perfdata = [] + infos = [] + status = 0 + for what, ctr in [ ("read", line[5]), ("write", line[8]) ]: + countername = "diskstat.%s.%s" % (item, what) + + # compute IO rate in bytes/sec + timedif, sectors_per_sec = get_counter(countername, this_time, int(ctr)) + bytes_per_sec = sectors_per_sec * 512 + infos.append("%s/sec %s" % (get_bytes_human_readable(bytes_per_sec), what)) + perfdata.append( (what, bytes_per_sec) ) + + # compute average of the rate over ___ minutes + if average_range != None: + timedif, avg = get_average(countername + ".avg", this_time, bytes_per_sec, average_range * 60) + perfdata.append( (what + ".avg", avg) ) + bytes_per_sec = avg + + # check levels + levels = params.get(what) + if levels != None: + mb_per_sec = bytes_per_sec / 1048576 + warn, crit = levels + if mb_per_sec >= crit: + status = 2 + infos[-1] += "!!" + elif mb_per_sec >= warn: + status = max(status, 1) + infos[-1] += "!" + + if average_range != None: + perfdata = [ perfdata[0], perfdata[2], perfdata[1], perfdata[3] ] - timedif, per_sec = get_counter("diskstat." + item, this_time, this_val) - mb_per_s = per_sec / 2048.0 # Diskstat output is in sectors a 512 Byte - perfdata = [ (item, "%dc" % this_val ) ] - return (0, "OK - %.1fMB/s (in last %d secs)" % (mb_per_s, timedif), perfdata) + return (status, nagios_state_names[status] + " - " + ", ".join(infos) , perfdata) + + return (3, "UNKNOWN - device missing") check_info['diskstat'] = (check_diskstat, "Disk IO %s", 1, inventory_diskstat) diff --git a/pnp-templates/check_mk-diskstat.php b/pnp-templates/check_mk-diskstat.php index 195c0e4..48b006d 100644 --- a/pnp-templates/check_mk-diskstat.php +++ b/pnp-templates/check_mk-diskstat.php @@ -23,12 +23,46 @@ # to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, # Boston, MA 02110-1301 USA. -$opt[1] = "--vertical-label 'Througput (MByte/s)' -l0 -u 1 --title \"Disk throughput $hostname / $servicedesc\" "; +$parts = explode("_", $servicedesc); +$disk = $parts[2]; -$def[1] = "DEF:kb=$RRDFILE[1]:$DS[1]:AVERAGE " ; -$def[1] .= "CDEF:mb=kb,1024,/ " ; -$def[1] .= "AREA:mb#40c080 " ; -$def[1] .= "GPRINT:mb:LAST:\"%6.1lf MByte/s last\" " ; -$def[1] .= "GPRINT:mb:AVERAGE:\"%6.1lf MByte/s avg\" " ; -$def[1] .= "GPRINT:mb:MAX:\"%6.1lf MByte/s max\\n\" "; +$opt[1] = "--vertical-label 'Throughput (MB/s)' -X0 --title \"Disk throughput $hostname / $disk\" "; + +$def[1] = + "HRULE:0#a0a0a0 ". +# read + "DEF:read=$RRDFILE[1]:$DS[1]:MAX ". + "CDEF:read_mb=read,1048576,/ ". + "AREA:read_mb#40c080:\"Read \" ". + "GPRINT:read_mb:LAST:\"%6.1lf MByte/s last\" ". + "GPRINT:read_mb:AVERAGE:\"%6.1lf MByte/s avg\" ". + "GPRINT:read_mb:MAX:\"%6.1lf MByte/s max\\n\" "; + +# read average +if (isset($DS[3])) { + $def[1] .= + "DEF:read_avg=$RRDFILE[3]:$DS[3]:MAX ". + "CDEF:read_avg_mb=read_avg,1048576,/ ". + "LINE:read_avg_mb#202020 "; +} + +# write +$def[1] .= + "DEF:write=$RRDFILE[2]:$DS[2]:MAX ". + "CDEF:write_mb=write,1048576,/ ". + "CDEF:write_mb_neg=write_mb,-1,* ". + "AREA:write_mb_neg#4080c0:Write ". + "GPRINT:write_mb:LAST:\"%6.1lf MByte/s last\" ". + "GPRINT:write_mb:AVERAGE:\"%6.1lf MByte/s avg\" ". + "GPRINT:write_mb:MAX:\"%6.1lf MByte/s max\\n\" ". + ""; + +# write average +if (isset($DS[3])) { + $def[1] .= + "DEF:write_avg=$RRDFILE[4]:$DS[4]:MAX ". + "CDEF:write_avg_mb=write_avg,1048576,/ ". + "CDEF:write_avg_mb_neg=write_avg_mb,-1,* ". + "LINE:write_avg_mb_neg#202020 "; +} ?> diff --git a/web/plugins/perfometer/check_mk.py b/web/plugins/perfometer/check_mk.py index f7be5aa..8aaa712 100644 --- a/web/plugins/perfometer/check_mk.py +++ b/web/plugins/perfometer/check_mk.py @@ -271,3 +271,16 @@ def perfometer_check_mk_uptime(row, check_command, perf_data): perfometers["check_mk-uptime"] = perfometer_check_mk_uptime perfometers["check_mk-snmp_uptime"] = perfometer_check_mk_uptime + +def perfometer_check_mk_diskstat(row, check_command, perf_data): + read_bytes = float(perf_data[0][1]) + write_bytes = float(perf_data[1][1]) + + text = "%-.2fM/s %-.2fM/s" % \ + (read_bytes / (1024*1024.0), write_bytes / (1024*1024.0)) + + return text, perfometer_logarithmic_dual( + read_bytes, "#60e0a0", write_bytes, "#60a0e0", 5000000, 10) + +perfometers["check_mk-diskstat"] = perfometer_check_mk_diskstat + diff --git a/web/plugins/views/perfometer.py b/web/plugins/views/perfometer.py index afd7b84..2fdac40 100644 --- a/web/plugins/views/perfometer.py +++ b/web/plugins/views/perfometer.py @@ -61,6 +61,34 @@ def perfometer_logarithmic(value, half_value, base, color): "</tr></table>" +# Dual logarithmic Perf-O-Meter +def perfometer_logarithmic_dual(value_left, color_left, value_right, color_right, half_value, base): + result = '<table><tr>' + for where, value, color in [ + ("left", value_left, color_left), + ("right", value_right, color_right) ]: + value = float(value) + if value == 0.0: + pos = 0 + else: + half_value = float(half_value) + h = math.log(half_value, base) # value to be displayed at 50% + pos = 25 + 10.0 * (math.log(value, base) - h) + if pos < 1: + pos = 1 + if pos > 49: + pos = 49 + + if where == "right": + result += perfometer_td(pos, color) + \ + perfometer_td(50 - pos, "white") + else: + result += perfometer_td(50 - pos, "white") + \ + perfometer_td(pos, color) + + return result + '</tr></table>' + + perfometer_plugins_dir = defaults.web_dir + "/plugins/perfometer" for fn in os.listdir(perfometer_plugins_dir): if fn.endswith(".py"):