Advanced Modelling: When a single SNMP variable isn't enough


Occasionally you will come across a device or a situation where collecting a single SNMP variable is insufficient, for example when two or more SNMP properties need to be combined to provide a meaningful measurement.

NMIS version 8.4.8G and later support modelling such scenarios using custom variables, or CVARs. With this mechanism you can temporarily capture up to 10 separate SNMP properties as a CVAR and define an arbitrarily complex expression (in perl) that transforms these CVARs into the one measurement that you want to collect and/or display.


Where and How to use CVARs

CVARs are supported

  • in the test and value expressions in the NMIS alert and threshold subsystem,
  • in calculate expressions in the general modelling subsystem,
  • and from NMIS version 8.6 on, also in control expressions evereywhere (in versions before that only a single CVAR  was supported in control).

To use CVARs you define the required CVARs as holding a previously specified SNMP variable at the beginning of one of the supported expressions; Subsequently you can then reference the CVAR value in the part of the expression that calculates the desired value to be used by NMIS.

An example scenario

The DS3 MIB defines a variety or error counters for DS3 circuits like "dsx3CurrentLCVs" which are based on a 15 minute observation interval and reset automatically at the end of the interval. As the interval start and end is arbitrary and up to the device to set, just capturing the error counters themselves is not quite workable. However, the DS3 MIB also specifies the variable "dsx3TimeElapsed" that holds the seconds elapsed since the start of the current observation interval. Dividing the raw error counter by the number of seconds into the interval results in a normalised errors-per-second rate which works well for collection and display.

Here is an excerpt of the relevant model file:

'systemHealth' => 
{
	'sections' => 'ds3Errors',
	'sys' => 
	{
		'ds3Errors' => 
		{        
			'indexed' => 'dsx3CurrentIndex',        
			'index_oid' => '1.3.6.1.2.1.10.30.6.1.1', 
			'headers' => 'ds3intf,ds3linestatus',
			'snmp' => {
				'ds3intf' => {
					'oid' => '1.3.6.1.2.1.2.2.1.2', # ifDescr
					'title' => 'DS3 Interface',
				},
				'ds3linestatus' => {
					'oid' => '1.3.6.1.2.1.10.30.5.1.10', # dsx3LineStatus
					'title' => "DS3 Line Status",
					'calculate' => 'my @x; my %triggers=(1,"No Alarm",2,"Rx Remote Alarm",4,"Tx Remote Alarm",8,"Rx AIS",16,"Tx AIS",32,"Rx LOF",64,"Rx LOS",128,"Loopback",256,"Test Pattern",512,"Unknown",1024,"Near end unavailable signal",2048,"Carrier Equip OOS"); while (my ($num,$txt)=each(%triggers)) { push (@x,$txt) if (int($r) & int($num)); }; return join(", ",@x); ',
				},
...
				'ds3LCV' => {
					'oid' => '1.3.6.1.2.1.10.30.6.1.6', # dsx3CurrentLCVs
					'title' => 'Line Coding Violations per second',
					'calculate' => 'CVAR1=ds3Elapsed; return ($CVAR1? $r/$CVAR1 : 0);',
		},
	},	# sys
				
	'rrd' => {
		'ds3Errors' => {
			'indexed' => 'true',
			"graphtype" => "ds3Errors",
			"snmp" => {
				'ds3Elapsed' => {
					'oid' => '1.3.6.1.2.1.10.30.5.1.3', # dsx3TimeElapsed
					'title' => 'elapsed seconds in current measurement interval',
					'option' => 'gauge,0:U',
				},
...
				"ds3LCV" => {
					'oid' => '1.3.6.1.2.1.10.30.6.1.6',
					'option' => 'gauge,0:U',
					'title' => "Line Coding Violations per second",
					'calculate' => 'CVAR1=ds3Elapsed; return ($CVAR1? $r/$CVAR1 : 0);',
				},
	}, 	# rrd
},	# systemhealth 

In the example above, the calculate expressions are used in two ways:

  • to transform the bitfield variable "DS3 Line Status" into a more verbose textual list of component statuses,
  • and to divide the raw dsx3CurrentLCVs error count by the dsx3TimeElapsed interval length.

In both cases the syntax is very straight-forward:

  • The expression must be a valid perl statement and return exactly one value.
  • The tokens $r, and CVAR0 to CVAR9 are interpreted by NMIS; everything else is perl.
  • Defining and using local variables with my is ok, but don't attempt to change any global NMIS variables.
  • "CVAR1=some_snmp_var;" defines what SNMP object CVAR1 is supposed to hold. The parser understands CVAR0 to CVAR9 for a total of 10 captures.
  • You can use functions that were defined elsewhere in NMIS in your calculate expression.
    You will likely have to include the full module namespace in the function call, e.g. func::beautify_physaddress(...).
    Only functions without side-effects should be used.
  • "return $r/$CVAR1;" accesses the value of CVAR1 in an expression. The variable "$r" represents the SNMP variable that the calculate expression is attached to.

Please note that

  • the $CVARn replacement in the expression is performed on a purely textual basis, before the expression is handed to the perl interpreter for evaluation :
    • For string variables you have to provide quotes in your expression, e.g.

      calculate => 'CVAR1=somestringthing; return 42 if ("$CVAR1" eq "online");'
    • Numeric variables can be used straight without quotes.
  • the $CVARn access refers to the raw value of the named property, ie. the data before any replace or calculate expressions for the named property were evaluated.

How to keep temporary CVAR data out of the RRD databases

As outlined above all the objects that you want to access via CVARs must be defined in the same section. If your test/calculate expression is within an rrd section, all the other objects will have to be within that rrd section, too, and thus they would be collected by NMIS and stored in RRD - quite wasteful if these other variables are just temporary and only there to for access using one CVAR expression.

In versions 8.6.0 and above you can prevent this by adding an option with value nosave:

'snmp' => {
  'hrNumUsers' => {
     'oid' => 'hrSystemNumUsers',
     'option' => 'nosave',
  },

In the example above,  hrNumUsers would be retrieved with SNMP, and other variables could be defined in terms of e.g. CVAR3=hrNumUsers, but hrNumUsers would not be saved.

Please note that setting nosave disables alerts for the given object.

Using Placeholders

Sometimes there is a need to create calculated data using data gathered from several sources and calculated into a single table.  Situations such as this might call for use of a plugin (NMIS 9 Collect and Update Plugins).  The problem comes in when the model tries to interpret the sections in the model when there is lack of data to do so.  This causes errors in the model.  To solve this, there is a keyword. 'placeholder' which can be used in the 'sys' section of the model definition to tell the model that the 'headers', labels, and 'titles' only exist to create the necessary infrastructure for the data, and the some other method will be used to create the values. There really isn't much to this, just the keyword 'placeholder', and the value will be printed in the log to say that the <value> will be used to fill in the data as shown in the example below.

      'ciscoNormalizedCPUMem' => {
        'headers' => 'TotalCPUs,MemoryUsedMax,MemoryUsed,MemoryFreeMax,MemoryFree',
        'placeholder' => 'plugin',
        'graphtype' => 'health',
        'indexed' => 'true',
        'snmp' => {
          'TotalCPUs' => {
            'title' => 'Number of CPUs'
          },
          'MemoryUsedMax' => {
            'title' => 'Maximum Memory Used'
          },
          'MemoryUsed' => {
            'title' => 'Current Memory Used'
          },
          'MemoryFreeMax' => {
            'title' => 'Maximum Memory Free'
          },
          'MemoryFree' => {
            'title' => 'Current Memory Free'
          },
        },
      },