C.chapman: /* Monitoring CPU load */

2010-03-08T16:22:43Z

‎Monitoring CPU load

Dts12: Wikification of HTML Tutorial page

2007-10-31T15:44:30Z

Wikification of HTML Tutorial page

New page

Broadly speaking, there are two basic techniques for monitoring a system using SNMP.
Either the agent can be configured to watch for certain values or events, and report
to a management station when something significant happens.
Alternatively, some external application can be set up that will poll the agent
regularly, and collect the information of interest.

The [[TUT:DisMan Monitoring|DisMan Monitoring]] page describes the first approach -
this tutorial will concentrate on the second, using the [http://www.mrtg.org Multi Router Traffic Grapher]
tool.

Note that this does not attempt to be a complete tutorial on the use of MRTG,
or even cover monitoring network traffic. Such topics are comprehensively
covered on the MRTG site already. This page is specifically concerned with
discussing other values that can be monitored using MRTG.

== Monitoring CPU load ==

There are several ways to try to monitor CPU usage, because the system maintains several CPU counters: Idle, System, User, Nice. As MRTG does not allow you to graph more than two values at a time, it is possible to select any two values ang graoh aginst each other, but it it also possible to compute a combined value.

=== User vs Idle CPU usage ===

The MRTG configuration would look something like

LoadMIBs: /usr/local/share/snmp/mibs/UCD-SNMP-MIB.txt

Target[kontor.cpu]:ssCpuRawUser.0&ssCpuRawIdle.0:public@10.10.11.2
RouterUptime[kontor.cpu]: public@10.10.11.2
MaxBytes[kontor.cpu]: 100
Title[kontor.cpu]: CPU LOAD
PageTop[kontor.cpu]: <H1>User CPU Load %</H1>
Unscaled[kontor.cpu]: ymwd
ShortLegend[kontor.cpu]: %
YLegend[kontor.cpu]: CPU Utilization
Legend1[kontor.cpu]: User CPU in % (Load)
Legend2[kontor.cpu]: Idle CPU in % (Load)
Legend3[kontor.cpu]:
Legend4[kontor.cpu]:
LegendI[kontor.cpu]: User
LegendO[kontor.cpu]: Idle
Options[kontor.cpu]: growright,nopercent

=== User vs System CPU usage ===

The MRTG configuration would look something like

LoadMIBs: /usr/local/share/snmp/mibs/UCD-SNMP-MIB.txt

Target[kontor.usrsys]:ssCpuRawUser.0&ssCpuRawSystem.0:public@10.10.11.2
RouterUptime[kontor.usrsys]: public@10.10.11.2
MaxBytes[kontor.usrsys]: 100
Title[kontor.usrsys]: CPU LOAD
PageTop[kontor.usrsys]: <H1>CPU (user and system) Load %</H1>
Unscaled[kontor.usrsys]: ymwd
ShortLegend[kontor.usrsys]: %
YLegend[kontor.usrsys]: CPU Utilization
Legend1[kontor.usrsys]: User CPU in % (Load)
Legend2[kontor.usrsys]: System CPU in % (Load)
Legend3[kontor.usrsys]:
Legend4[kontor.usrsys]:
LegendI[kontor.usrsys]: User
LegendO[kontor.usrsys]: System
Options[kontor.usrsys]: growright,nopercent

=== Active CPU usage ===

This example sums User, System and Nice and plots it as a single value.

The corresponding MRTG configuration is

LoadMIBs: /usr/local/share/snmp/mibs/UCD-SNMP-MIB.txt

Target[kontor.cpusum]:ssCpuRawUser.0&ssCpuRawUser.0:public@10.10.11.2 + \
ssCpuRawSystem.0&ssCpuRawSystem.0:public@10.10.11.2 + \
ssCpuRawNice.0&ssCpuRawNice.0:public@10.10.11.2
RouterUptime[kontor.cpusum]: public@10.10.11.2
MaxBytes[kontor.cpusum]: 100
Title[kontor.cpusum]: CPU LOAD
PageTop[kontor.cpusum]: <H1>Active CPU Load %</H1>
Unscaled[kontor.cpusum]: ymwd
ShortLegend[kontor.cpusum]: %
YLegend[kontor.cpusum]: CPU Utilization
Legend1[kontor.cpusum]: Active CPU in % (Load)
Legend2[kontor.cpusum]:
Legend3[kontor.cpusum]:
Legend4[kontor.cpusum]:
LegendI[kontor.cpusum]: Active
LegendO[kontor.cpusum]:
Options[kontor.cpusum]: growright,nopercent

== Monitoring Disk space ==

There are two tables that can be used when monitoring disk space, UCD-SNMP-MIB:dskTable and HOST-RESOURCES-MIB:hrStorageTable. There are two noticeable differences bwtween these tables: UCD-SNMP-MIB:dskTable requires a configuration directive in snmpd.conf, but it also gives a direct measure of the percentage of free space

On the system I used for these examples, I have the following disks:

$ df
Filesystem Size Used Avail Use% Mounted on
/dev/hda6 289M 56M 218M 21% /
/dev/hda1 7.3G 3.2G 4.1G 44% /mnt/windows
/dev/hda7 1.7G 1.2G 429M 74% /usr
/dev/hdc1 19G 6.7G 11G 37% /space

and I have put the following lines into snmpd.conf:

disk / 100000
disk /usr 100000
disk /space 100000

=== Monitoring dskTable ===

This plots the usage of the root and /usr partitions in one graph

LoadMIBs: /usr/local/share/snmp/mibs/UCD-SNMP-MIB.txt

Target[kontor.root]:dskPercent.1&dskPercent.2:public@10.10.11.2
RouterUptime[kontor.root]: public@10.10.11.2
MaxBytes[kontor.root]: 100
Title[kontor.root]: DISK USAGE
PageTop[kontor.root]: <H1>DISK / and /usr Usage %</H1>
Unscaled[kontor.root]: ymwd
ShortLegend[kontor.root]: %
YLegend[kontor.root]: DISK Utilization
Legend1[kontor.root]: Root disk
Legend2[kontor.root]: /usr disk
Legend3[kontor.root]:
Legend4[kontor.root]:
LegendI[kontor.root]: Root disk
LegendO[kontor.root]: /usr disk
Options[kontor.root]: growright,gauge,nopercent

=== Monitoring hrStorageTable ===

This plots the partition size as one value, and the used space as another.

Note: the values returned here are in units of hrStorageUnits, which, dependent on the file system type, may or may not be 1K.

LoadMIBs: /usr/local/share/snmp/mibs/HOST-RESOURCES-MIB.txt

Target[kontor.hrroot]:hrStorageSize.1&hrStorageUsed.1:public@10.10.11.2
RouterUptime[kontor.hrroot]: public@10.10.11.2
MaxBytes[kontor.hrroot]: 300000
Title[kontor.hrroot]: DISK / USAGE
PageTop[kontor.hrroot]: <H1>DISK / Usage</H1>
ShortLegend[kontor.hrroot]: B
kMG[kontor.hrroot]: k,M,G,T,P
kilo[kontor.hrroot]: 1024
YLegend[kontor.hrroot]: DISK / Utilization
Legend1[kontor.hrroot]: Root disk size
Legend2[kontor.hrroot]: Root disk usage
Legend3[kontor.hrroot]:
Legend4[kontor.hrroot]:
LegendI[kontor.hrroot]: Root disk size
LegendO[kontor.hrroot]: Root disk usage
Options[kontor.hrroot]: growright,gauge,nopercent

== Other Monitoring Examples ==

Two further examples that have been offered:

=== Monitoring TCP connections ===

Target[tcpopen]:
.1.3.6.1.2.1.6.9.0&.1.3.6.1.2.1.6.9.0:public@localhost
Options[tcpopen]: nopercent,growright,gauge,noinfo
Title[tcpopen]: Open TCP connections
PageTop[tcpopen]: Open TCP connections
MaxBytes[tcpopen]: 1000000
YLegend[tcpopen]: # conns
ShortLegend[tcpopen]: connections
LegendI[tcpopen]: Connections:
LegendO[tcpopen]:
Legend1[tcpopen]: Open TCP connections

=== Monitoring Free Memory ===

Target[freemem]:
.1.3.6.1.4.1.2021.4.11.0&.1.3.6.1.4.1.2021.4.11.0:public@localhost
Options[freemem]: nopercent,growright,gauge,noinfo
Title[freemem]: Free Memory
PageTop[freemem]: Free Memory
MaxBytes[freemem]: 1000000
kMG[freemem]: k,M,G,T,P,X
YLegend[freemem]: bytes
ShortLegend[freemem]: bytes
LegendI[freemem]: Free Memory:
LegendO[freemem]:
Legend1[freemem]: Free memory, not including swap, in bytes

<hr>

{{TUT:LIST}}

@@ Line 16: / Line 16: @@
 == Monitoring CPU load ==
-There are several ways to try to monitor CPU usage, because the system maintains several CPU counters: Idle, System, User, Nice. As MRTG does not allow you to graph more than two values at a time, it is possible to select any two values ang graoh aginst each other, but it it also possible to compute a combined value.
+There are several ways to try to monitor CPU usage, because the system maintains several CPU counters: Idle, System, User, Nice. As MRTG does not allow you to graph more than two values at a time, it is possible to select any two values and graph against each other.  It is also possible to compute a combined value.
 === User vs Idle CPU usage ===

TUT:MRTG - Revision history

C.chapman: /* Monitoring CPU load */

Dts12: Wikification of HTML Tutorial page