introduce averaging Message-ID: <516bbec2.L4Ldid4nyDYj/UnB%mk(a)mathias-kettner.de> User-Agent: Heirloom mailx 12.4 7/29/08 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Module: check_mk Branch: master Commit: caf4f1d640a5aaedfa10403ba32da205aa175596 URL: http://git.mathias-kettner.de/git/?p=check_mk.git;a=commit;h=caf4f1d640a5aa… Author: Mathias Kettner <mk(a)mathias-kettner.de> Date: Mon Apr 15 10:47:58 2013 +0200 esx_vsphere_hostsystem.cpu_usage: introduce averaging --- checkman/esx_vsphere_hostsystem.cpu_usage | 21 ++++--- checks/esx_vsphere_hostsystem | 43 ++++++------- .../check_mk-esx_vsphere_hostsystem.cpu_usage.php | 63 ++++++++++++++------ web/plugins/perfometer/check_mk.py | 4 +- 4 files changed, 78 insertions(+), 53 deletions(-) diff --git a/checkman/esx_vsphere_hostsystem.cpu_usage b/checkman/esx_vsphere_hostsystem.cpu_usage index be957b4..d90f252 100644 --- a/checkman/esx_vsphere_hostsystem.cpu_usage +++ b/checkman/esx_vsphere_hostsystem.cpu_usage @@ -8,19 +8,24 @@ description: It also shows you the number of sockets, cores and threads. perfdata: - One value: the current usage in terms of cores. The maximum value sent - is the number of core. If you have 12 cores installed and the usage is 50%, - then a value of 6 will be sent. That way the performance graph is able - to show the number of cores correctly. The warn/crit levels are also - converted in relation to the number of cores. + One or two values: the first value is current usage in percent - ranging from + 0 to 100. The "maximum" value is not 100, but the number of CPU threads. This + case be used for scaling the graph in terms of the number of used CPU threads. + + If averaging is enabled then a second value is sent: the averaged CPU utilization + ranging from 0 to 100. inventory: One check per ESX Host System will be created. [parameters] -warning (float): A percentage for the warning level ranging from 0 to 100 -critical (float): A percentage for the critical level ranging from 0 to 100 +parameters(dict): A dictionary with the following keys: + + {"levels"}: Either {None} for no levels, a tuple of warn and crit (in percent) or + a dictionary with predictive levels settings. + + {"average"}: A number of minutes for enabling averaging. [configuration] -esx_host_cpu_default_levels(float, float): Default levels, preset to {None}, which means that no levels +esx_host_cpu_default_levels(dict): Default levels, preset to an empty dictionary, which means that no levels will be applied. diff --git a/checks/esx_vsphere_hostsystem b/checks/esx_vsphere_hostsystem index 65b4ab0..3f10e9c 100644 --- a/checks/esx_vsphere_hostsystem +++ b/checks/esx_vsphere_hostsystem @@ -41,7 +41,7 @@ def esx_vsphere_hostsystem_convert(info): # | | # +----------------------------------------------------------------------+ -esx_host_cpu_default_levels = None +esx_host_cpu_default_levels = {} # hardware.cpuInfo.numCpuCores 12 # hardware.cpuInfo.numCpuPackages 2 @@ -55,12 +55,12 @@ def inventory_esx_vsphere_hostsystem_cpu(info): if 'summary.quickStats.overallCpuUsage' in data \ and 'hardware.cpuInfo.hz' in data\ and 'hardware.cpuInfo.numCpuCores' in data: - return [(None, 'esx_host_cpu_default_levels')] + return [(None, {})] def check_esx_vsphere_hostsystem_cpu(item, params, info): data = esx_vsphere_hostsystem_convert(info) - num_cores = int(data['hardware.cpuInfo.numCpuCores'][0]) num_sockets = int(data['hardware.cpuInfo.numCpuPackages'][0]) + num_cores = int(data['hardware.cpuInfo.numCpuCores'][0]) num_threads = int(data['hardware.cpuInfo.numCpuThreads'][0]) used_mhz = float(data['summary.quickStats.overallCpuUsage'][0]) mhz_per_core = float(data['hardware.cpuInfo.hz'][0]) / 1024.0 / 1024.0 @@ -71,35 +71,30 @@ def check_esx_vsphere_hostsystem_cpu(item, params, info): infotext = "%.1f%%" % usage - state = 0 - if params: - warn, crit = params - if usage >= crit: - state = 2 - infotext += " (levels at %d%%/%d%%)" % (warn, crit) - elif usage >= warn: - state = 1 - infotext += " (levels at %d%%/%d%%)" % (warn, crit) - warn_perf = warn * per_core - crit_perf = crit * per_core - else: - warn_perf, crit_perf = None, None + # Convert legacy parameters + this_time = time.time() + state, infotext, perfdata = check_cpu_util(usage, params) infotext += ", %.2fGHz/%.2fGHz" % (used_mhz / 1024.0, total_mhz / 1024.0) - infotext += ", %d sockets, %d cores/socket, %d threads" % ( num_sockets, num_cores / num_sockets, num_threads) - perf = [("usage", usage * per_core, warn_perf, crit_perf, 0, 100 * per_core)] - return (state, infotext, perf) + # put number of threads as MAX value for first perf-data. This + # is needed by the PNP template. + perfdata_cpu = list(perfdata[0]) + perfdata_cpu[-1] = num_threads + perfdata = [ tuple(perfdata_cpu) ] + perfdata[1:] + return (state, infotext, perfdata) check_info['esx_vsphere_hostsystem.cpu_usage'] = { - "inventory_function" : inventory_esx_vsphere_hostsystem_cpu, - "check_function" : check_esx_vsphere_hostsystem_cpu, - "service_description" : "CPU utilization", - "group" : "cpu_utilization", - "has_perfdata" : True + "inventory_function" : inventory_esx_vsphere_hostsystem_cpu, + "check_function" : check_esx_vsphere_hostsystem_cpu, + "service_description" : "CPU utilization", + "group" : "cpu_utilization_os", + "has_perfdata" : True, + "default_levels_variable" : "esx_host_cpu_default_levels", + "includes" : [ "cpu_util.include" ], } diff --git a/pnp-templates/check_mk-esx_vsphere_hostsystem.cpu_usage.php b/pnp-templates/check_mk-esx_vsphere_hostsystem.cpu_usage.php index f4c47e3..82c4d1d 100644 --- a/pnp-templates/check_mk-esx_vsphere_hostsystem.cpu_usage.php +++ b/pnp-templates/check_mk-esx_vsphere_hostsystem.cpu_usage.php @@ -23,30 +23,57 @@ # to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, # Boston, MA 02110-1301 USA. -$num_cores = $MAX[1]; -$warnperc = $WARN[1] / $num_cores * 100.0; -$critperc = $CRIT[1] / $num_cores * 100.0; +# Do not depend on numbers, use names +$RRD = array(); +foreach ($NAME as $i => $n) { + $RRD[$n] = "$RRDFILE[$i]:$DS[$i]:MAX"; + $WARN[$n] = $WARN[$i]; + $CRIT[$n] = $CRIT[$i]; + $MIN[$n] = $MIN[$i]; + $MAX[$n] = $MAX[$i]; +} + +$num_threads = $MAX[1]; +$warnthreads = $WARN[1] * $num_threads / 100.0; +$critthreads = $CRIT[1] * $num_threads / 100.0; +$rightscale = 100.0 / $num_threads; + +$opt[1] = "--vertical-label 'Used CPU threads' --right-axis $rightscale:0 --right-axis-format '%4.1lf%%' -l0 -ru $num_threads --title \"CPU Utilization for $hostname ($num_threads CPU threads)\" "; -$opt[1] = "--vertical-label 'Used cores' -l0 -ru $num_cores --title \"CPU Utilization for $hostname\" "; +$def[1] = "DEF:perc=$RRD[util] " + . "CDEF:util=perc,$num_threads,*,100,/ " + ; + +$def[1] .= "HRULE:$MAX[util]#0040d0:\"$num_threads CPU Threads\\n\" " + ; -$def[1] = "DEF:util=$RRDFILE[1]:$DS[1]:MAX " - . "CDEF:perc=util,$num_cores,/,100,* " - . "AREA:util#60f020:\"Utilization\:\" " - . "LINE:util#308010 " - . "GPRINT:perc:LAST:\"%0.1lf%% \" "; +$def[1] .= "AREA:util#60f020:\"Utilization\:\" " + . "LINE:util#50b01a " + . "GPRINT:perc:LAST:\"%.1lf%%\" " + . "GPRINT:util:LAST:\"(%.1lf Threads) \" " + . "GPRINT:perc:MIN:\"min\: %.1lf%%,\" " + . "GPRINT:util:MIN:\"(%.1lf), \" " + . "GPRINT:perc:MAX:\"max\: %.1lf%%\" " + . "GPRINT:util:MAX:\"(%.1lf)\\n\" " + ; -if ($WARN[1]) { - $def[1] .= "HRULE:$WARN[1]#fff000:\"Warn at $warnperc% \" " - . "HRULE:$CRIT[1]#ff0000:\"Critical at $critperc%\\n\" "; + +if (isset($RRD["avg"])) { + $def[1] .= "DEF:aperc=$RRD[avg] ". + "CDEF:avg=aperc,$num_threads,*,100,/ ". + "LINE:avg#004000:\"Averaged\: \" ". + "GPRINT:aperc:LAST:\"%.1lf%%,\" ". + "GPRINT:aperc:MIN:\"min\: %.1lf%%,\" ". + "GPRINT:aperc:MAX:\"max\: %.1lf%%\\n\" ". + ""; +} + +if ($WARN['util']) { + $def[1] .= "HRULE:$warnthreads#fff000:\"Warn at $WARN[util]% \" " + . "HRULE:$critthreads#ff0000:\"Critical at $CRIT[util]%\\n\" "; } else { $def[1] .= "COMMENT:\"\\n\" "; } -$def[1] .= "HRULE:$MAX[1]#0040d0:\"$num_cores Cores installed \" " - . "GPRINT:util:MIN:\"Min\: %5.2lf Cores \" " - . "GPRINT:util:MAX:\"Max\: %5.2lf Cores\" " - . "GPRINT:util:LAST:\"Last\: %4.1lf Cores\\n\" " - ; - ?> diff --git a/web/plugins/perfometer/check_mk.py b/web/plugins/perfometer/check_mk.py index f959cf6..fd6290d 100644 --- a/web/plugins/perfometer/check_mk.py +++ b/web/plugins/perfometer/check_mk.py @@ -646,9 +646,7 @@ perfometers['check_mk-esx_vsphere_hostsystem.mem_usage'] = perfometer_simple_mem perfometers['check_mk-esx_vsphere_virtualmachine.mem_usage'] = perfometer_simple_mem_usage def perfometer_esx_vsphere_hostsystem_cpu(row, command, perf): - cores = float(perf[0][6]) - used = float(perf[0][1]) - used_perc = used / cores * 100.0 + used_perc = float(perf[0][1]) return "%d%%" % used_perc, perfometer_linear(used_perc, "#60f020") perfometers['check_mk-esx_vsphere_hostsystem.cpu_usage'] = perfometer_esx_vsphere_hostsystem_cpu