The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


SNMP - The Perl5 'SNMP' Extension Module for the Net-SNMP SNMP package.


 use SNMP;
 $sess = new SNMP::Session(DestHost => localhost, Community => public);
 $val = $sess->get('sysDescr.0');
 $vars = new SNMP::VarList([sysDescr,0], [sysContact,0], [sysLocation,0]);
 @vals = $sess->get($vars);
 $vb = new SNMP::Varbind();
 do {
    $val = $sess->getnext($vb);
    print "@{$vb}\n";
 } until ($sess->{ErrorNum});
 $SNMP::save_descriptions = 1;
 SNMP::initMib(); # assuming mib is not already loaded
 print "$SNMP::MIB{sysDescr}{description}\n";


Note: The perl SNMP 5.0 module which comes with net-snmp 5.0 and higher is different than previous versions in a number of ways. Most importantly, it behaves like a proper net-snmp application and calls init_snmp properly, which means it will read configuration files and use those defaults where appropriate automatically parse MIB files, etc. This will likely affect your perl applications if you have, for instance, default values set up in your snmp.conf file (as the perl module will now make use of those defaults). The docmuentation, however, has sadly not been updated yet (aside from this note), nor is the read_config default usage implementation fully complete.

The basic operations of the SNMP protocol are provided by this module through an object oriented interface for modularity and ease of use. The primary class is SNMP::Session which encapsulates the persistent aspects of a connection between the management application and the managed agent. Internally the class is implemented as a blessed hash reference. This class supplies 'get', 'getnext', 'set', 'fget', and 'fgetnext' method calls. The methods take a variety of input argument formats and support both syncronous and asyncronous operation through a polymorphic API (i.e., method behaviour varies dependent on args passed - see below).


$sess = new SNMP::Session(DestHost => 'host', ...)

The following arguments may be passed to new as a hash.


default 'localhost', hostname or ip addr of SNMP agent


default 'public', SNMP community string (used for both R/W)


default taken from library configuration - probably 3 [1, 2 (same as 2c), 2c, 3]


default '161', allow remote UDP port to be overriden


default '1000000', micro-seconds before retry


default '5', retries before failure


default '0', if enabled NOSUCH errors in 'get' pdus will be repaired, removing the varbind in error, and resent - undef will be returned for all NOSUCH varbinds, when set to '0' this feature is disabled and the entire get request will fail on any NOSUCH error (applies to v1 only)


default 'initial', security name (v3)


default 'noAuthNoPriv', security level [noAuthNoPriv, authNoPriv, authPriv] (v3)


default <none>, security engineID, will be probed if not supplied (v3)


default <SecEngineId>, context engineID, will be probed if not supplied (v3)


default '', context name (v3)


default 'MD5', authentication protocol [MD5, SHA] (v3)


default <none>, authentication passphrase


default 'DES', privacy protocol [DES, AES] (v3)


default <none>, privacy passphrase (v3)


Directly specified SNMPv3 USM user keys (used if you want to specify the keys instead of deriving them from a password as above).


default 'undef', used by 'fget[next]', holds an hash reference of output value formatters, (e.g., {<obj> => <sub-ref>, ... }, <obj> must match the <obj> and format used in the get operation. A special <obj>, '*', may be used to apply all <obj>s, the supplied sub is called to translate the value to a new format. The sub is called passing the Varbind as the arg


default 'undef', used by 'fget[next]', holds an hash reference of output value formatters, (e.g., {<type> => <sub-ref>, ... }, the supplied sub is called to translate the value to a new format, unless a VarFormat mathces first (e.g., $sess->{TypeFormats}{INTEGER} = \&mapEnum(); although this can be done more efficiently by enabling $SNMP::use_enums or session creation param 'UseEnums')


defaults to the value of SNMP::use_long_names at time of session creation. set to non-zero to have <tags> for 'getnext' methods generated preferring longer Mib name convention (e.g., system.sysDescr vs just sysDescr)


defaults to the value of SNMP::use_sprint_value at time of session creation. set to non-zero to have return values for 'get' and 'getnext' methods formatted with the libraries snprint_value function. This will result in certain data types being returned in non-canonical format Note: values returned with this option set may not be appropriate for 'set' operations (see discussion of value formats in <vars> description section)


defaults to the value of SNMP::use_enums at time of session creation. set to non-zero to have integer return values converted to enumeration identifiers if possible, these values will also be acceptable when supplied to 'set' operations


defaults to the value of SNMP::use_numeric at time of session creation. set to non-zero to have <tags> for get methods returned as numeric OID's rather than descriptions. UseLongNames will be set so that the full OID is returned to the caller.


defaults to the value of SNMP::best_guess at time of session creation. this setting controls how <tags> are parsed. setting to 0 causes a regular lookup. setting to 1 causes a regular expression match (defined as -Ib in snmpcmd) and setting to 2 causes a random access lookup (defined as -IR in snmpcmd).


defaults to the value of SNMP::non_increasing at time of session creation. this setting controls if a non-increasing OID during bulkwalk will causes an error. setting to 0 causes the default behaviour (which may, in very badly performing agents, result in a never-ending loop). setting to 1 causes an error (OID not increasing) when this error occur.


read-only, holds the error message assoc. w/ last request


read-only, holds the snmp_err or staus of last request


read-only, holds the snmp_err_index when appropriate

Private variables:


internal field used to hold the translated DestHost field


internal field used to cache a created session structure

SNMP::Session methods


Updates the SNMP::Session object with the values fields passed in as a hash list (similar to new(<fields>)) (WARNING! not fully implemented)

$sess->get(<vars> [,<callback>])

do SNMP GET, multiple <vars> formats accepted. for syncronous operation <vars> will be updated with value(s) and type(s) and will also return retrieved value(s). If <callback> supplied method will operate asyncronously

$sess->fget(<vars> [,<callback>])

do SNMP GET like 'get' and format the values according the handlers specified in $sess->{VarFormats} and $sess->{TypeFormats}

$sess->getnext(<vars> [,<callback>])

do SNMP GETNEXT, multiple <vars> formats accepted, returns retrieved value(s), <vars> passed as arguments are updated to indicate next lexicographical <obj>,<iid>,<val>, and <type>

Note: simple string <vars>,(e.g., 'sysDescr.0') form is not updated. If <callback> supplied method will operate asyncronously

$sess->fgetnext(<vars> [,<callback>])

do SNMP GETNEXT like getnext and format the values according the handlers specified in $sess->{VarFormats} and $sess->{TypeFormats}

$sess->set(<vars> [,<callback>])

do SNMP SET, multiple <vars> formats accepted. the value field in all <vars> formats must be in a canonical format (i.e., well known format) to ensure unambiguous translation to SNMP MIB data value (see discussion of canonical value format <vars> description section), returns snmp_errno. If <callback> supplied method will operate asyncronously

$sess->getbulk(<non-repeaters>, <max-repeaters>, <vars>)

do an SNMP GETBULK, from the list of Varbinds, the single next lexico instance is fetched for the first n Varbinds as defined by <non-repeaters>. For remaining Varbinds, the m lexico instances are retrieved each of the remaining Varbinds, where m is <max-repeaters>.

$sess->bulkwalk(<non-repeaters>, <max-repeaters>, <vars> [,<callback>])

Do a "bulkwalk" of the list of Varbinds. This is done by sending a GETBULK request (see getbulk() above) for the Varbinds. For each requested variable, the response is examined to see if the next lexico instance has left the requested sub-tree. Any further instances returned for this variable are ignored, and the walk for that sub-tree is considered complete.

If any sub-trees were not completed when the end of the responses is reached, another request is composed, consisting of the remaining variables. This process is repeated until all sub-trees have been completed, or too many packets have been exchanged (to avoid loops).

The bulkwalk() method returns an array containing an array of Varbinds, one for each requested variable, in the order of the variable requests. Upon error, bulkwalk() returns undef and sets $sess->ErrorStr and $sess->ErrorNum. If a callback is supplied, bulkwalk() returns the SNMP request id, and returns immediately. The callback will be called with the supplied argument list and the returned variables list.

Note: Because the client must "discover" that the tree is complete by comparing the returned variables with those that were requested, there is a potential "gotcha" when using the max-repeaters value. Consider the following code to print a list of interfaces and byte counts:

    $numInts = $sess->get('ifNumber.0');
    ($desc, $in, $out) = $sess->bulkwalk(0, $numInts,
                  [['ifDescr'], ['ifInOctets'], ['ifOutOctets']]);

    for $i (0..($numInts - 1)) {
        printf "Interface %4s: %s inOctets, %s outOctets\n",
                  $$desc[$i]->val, $$in[$i]->val, $$out[$i]->val;

This code will produce *two* requests to the agent -- the first to get the interface values, and the second to discover that all the information was in the first packet. To get around this, use '$numInts + 1' for the max_repeaters value. This asks the agent to include one additional (unrelated) variable that signals the end of the sub-tree, allowing bulkwalk() to determine that the request is complete.

$results = $sess->gettable(<TABLE OID>, <OPTIONgt)

This will retrieve an entire table of data and return a hash reference to that data. The returned hash reference will have indexes of the OID suffixes for the index data as the key. The value for each entry will be another hash containing the data for a given row. The keys to that hash will be the column names, and the values will be the data.



  use SNMP;
  use Data::Dumper;

  my $s = new SNMP::Session(DestHost => 'localhost');

  print Dumper($s->gettable('ifTable'));

On my machine produces:

  $VAR1 = {
            '6' => {
                     'ifMtu' => '1500',
                     'ifPhysAddress' => 'PV',
                     # ...
                     'ifInUnknownProtos' => '0'
            '4' => {
                     'ifMtu' => '1480',
                     'ifPhysAddress' => '',
                     # ...
                     'ifInUnknownProtos' => '0'
            # ...

By default, it will try to do as optimized retrieval as possible. It'll request multiple columns at once, and use GETBULK if possible. A few options may be specified by passing in an OPTIONS hash containing various parameters:

noindexes => 1

Instructs the code not to parse the indexes and place the results in the second hash. If you don't need the index data, this will be faster.

columns => [ colname1, ... ]

This specifies which columns to collect. By default, it will try to collect all the columns defined in the MIB table.

repeat => COUNT

Specifies a GETBULK repeat COUNT. IE, it will request this many varbinds back per column when using the GETBULK operation. Shortening this will mean smaller packets which may help going through some systems. By default, this value is calculated and attepmts to guess at what will fit all the results into 1000 bytes. This calculation is fairly safe, hopefully, but you can either raise or lower the number using this option if desired. In lossy networks, you want to make sure that the packets don't get fragmented and lowering this value is one way to help that.

nogetbulk => 1

Force the use of GETNEXT rather than GETBULK. (always true for SNMPv1, as it doesn't have GETBULK anyway). Some agents are great implementers of GETBULK and this allows you to force the use of GETNEXT oprations instead.

callback => \&subroutine
callback => [\&subroutine, optarg1, optarg2, ...]

If a callback is specified, gettable will return quickly without returning results. When the results are finally retrieved the callback subroutine will be called (see the other sections defining callback behaviour and how to make use of SNMP::MainLoop which is required fro this to work). An additional argument of the normal hash result will be added to the callback subroutine arguments.

Note 1: internally, the gettable function uses it's own callbacks which are passed to getnext/getbulk as appropriate.

Note 2: callback support is only available in the SNMP module version 5.04 and above. To test for this in code intending to support both versions prior to 5.04 and and 5.04 and up, the following should work:

  if ($response = $sess->gettable('ifTable', callback => \&my_sub)) {
      # got a response, gettable doesn't support callback
      $no_mainloop = 1;

Deciding on whether to use SNMP::MainLoop is left as an excersize to the reader since it depends on whether your code uses other callbacks as well.


$sess = new SNMP::Session(DestHost => 'host', ...)

supports all applicable fields from SNMP::Session (see above)

SNMP::TrapSession methods

$sess->trap(enterprise, agent, generic, specific, uptime, <vars>)
    $sess->trap(enterprise=>'.', # or 'ucdavis' [default]
                agent => '', # or 'localhost',[dflt 1st intf on host]
                generic => specific,  # can be omitted if 'specific' supplied
                specific => 5,        # can be omitted if 'generic' supplied
                uptime => 1234,       # dflt to localhost uptime (0 on win32)
                [[ifIndex, 1, 1],[sysLocation, 0, "here"]]); # optional vars
                                                             # always last
trap(oid, uptime, <vars>) - v2 format
    $sess->trap(oid => 'snmpRisingAlarm',
                uptime => 1234,
                [[ifIndex, 1, 1],[sysLocation, 0, "here"]]); # optional vars
                                                             # always last

Acceptable variable formats:

<vars> may be one of the following forms:


represents an array of MIB objects to get or set, implemented as a blessed reference to an array of SNMP::Varbinds, (e.g., [<varbind1>, <varbind2>, ...])


represents a single MIB object to get or set, implemented as a blessed reference to a 4 element array; [<obj>, <iid>, <val>, <type>].


one of the following forms:


leaf identifier (e.g., 'sysDescr') assumed to be unique for practical purposes


fully qualified identifier (e.g., '')


fully qualified, dotted-decimal, numeric OID (e.g., '.')


the dotted-decimal, instance identifier. for scalar MIB objects use '0'


the SNMP data value retrieved from or being set to the agents MIB. for (f)get(next) operations <val> may have a variety of formats as determined by session and package settings. However for set operations the <val> format must be canonical to ensure unambiguous translation. The canonical forms are as follows:


dotted-decimal (e.g., .


perl scalar containing octets


decimal signed integer (or enum)






decimal unsigned integer


decimal unsigned integer


decimal unsigned integer


decimal unsigned integer


decimal unsigned integer


perl scalar containing octets


perl scalar containing nothing


SNMP data type (see list above), this field is populated by 'get' and 'getnext' operations. In some cases the programmer needs to populate this field when passing to a 'set' operation. this field need not be supplied when the attribute indicated by <tag> is already described by loaded Mib modules. for 'set's, if a numeric OID is used and the object is not currently in the loaded Mib, the <type> field must be supplied

simple string

light weight form of <var> used to 'set' or 'get' a single attribute without constructing an SNMP::Varbind. stored in a perl scalar, has the form '<tag>.<iid>', (e.g., 'sysDescr.0'). for 'set' operations the value is passed as a second arg. Note: This argument form is not updated in get[next] operations as are the other forms.

Acceptable callback formats

<callback> may be one of the following forms:

without arguments
sub { ... }
or with arguments
[ \&subname, $arg1, ... ]
[ sub { ... }, $arg1, ... ]
[ "method", $obj, $arg1, ... ]

callback will be called when response is received or timeout occurs. the last argument passed to callback will be a SNMP::VarList reference. In case of timeout the last argument will be undef.

&SNMP::MainLoop([<timeout>, [<callback>]])

to be used with async SNMP::Session calls. MainLoop must be called after initial async calls so return packets from the agent will not be processed. If no args suplied this function enters an infinite loop so program must be exited in a callback or externally interupted. If <timeout(sic)


This function, when called from an SNMP::MainLoop() callback function, will cause the current SNMP::MainLoop() to return after the callback is completed. finish() can be used to terminate an otherwise-infinite MainLoop. A new MainLoop() instance can then be started to handle further requests.

SNMP package variables and functions


the current version specifier (e.g., 3.1.0)


default '1', set to 0 to disable automatic reading of the MIB upon session creation. set to non-zero to call initMib at session creation which will result in MIB loading according to Net-SNMP env. variables (see man mib_api)


default '0', controls warning/info output of SNMP module, 0 => no output, 1 => enables warning/info output from SNMP module itself (is also controlled by SNMP::debugging - see below)


default '0', set to non-zero to enable the use of longer Mib identifiers. see translateObj. will also influence the formatting of <tag> in varbinds returned from 'getnext' operations. Can be set on a per session basis (UseLongNames)


default '0', set to non-zero to enable formatting of response values using the snmp libraries snprint_value function. can also be set on a per session basis (see UseSprintValue) Note: returned values may not be suitable for 'set' operations


default '0',set non-zero to return values as enums and allow sets using enums where appropriate. integer data will still be accepted for set operations. can also be set on a per session basis (see UseEnums)


default to '0',set to non-zero to have <tags> for 'get' methods returned as numeric OID's rather than descriptions. UseLongNames will be set so that the entire OID will be returned. Set on a per-session basis (see UseNumeric).


default '0'. This setting controls how <tags> are parsed. Setting to 0 causes a regular lookup. Setting to 1 causes a regular expression match (defined as -Ib in snmpcmd) and setting to 2 causes a random access lookup (defined as -IR in snmpcmd). Can also be set on a per session basis (see BestGuess)


default '0',set non-zero to have mib parser save attribute descriptions. must be set prior to mib initialization


default '0', controlls debugging output level within SNMP module and libsnmp

  1. enables 'SNMP::verbose' (see above)

  2. level 1 plus snmp_set_do_debugging(1)

  3. level 2 plus snmp_set_dump_packet(1)


default '0', set [non-]zero to independently set snmp_set_dump_packet()


a tied hash to access parsed MIB information. After the MIB has been loaded this hash allows access to to the parsed in MIB meta-data(the structure of the MIB (i.e., schema)). The hash returns blessed references to SNMP::MIB::NODE objects which represent a single MIB attribute. The nodes can be fetched with multiple 'key' formats - the leaf name (e.g.,sysDescr) or fully/partially qualified name (e.g., system.sysDescr) or fully qualified numeric OID. The returned node object supports the following fields:


dotted decimal fully qualified OID


leaf textual identifier (e.g., 'sysDescr')


leaf numeric OID component of objectID (e.g., '1')


textual identifier for module (e.g., 'RFC1213-MIB')


parent node


array reference of children nodes


next lexico node (BUG!does not return in lexico order)


returns application type (see getType for values)


returns ACCESS (ReadOnly, ReadWrite, WriteOnly, NoAccess, Notify, Create)


returns STATUS (Mandatory, Optional, Obsolete, Deprecated)


returns 'textualConvention' if defined else 'type'






returns UNITS


returns HINT


returns hash ref {tag => num, ...}


returns array ref of hash ref [{low => num, high => num}, ...]


returns DESCRIPTION ($SNMP::save_descriptions must be set prior to MIB initialization/parsing)


returns the REFERENCE clause

MIB Functions


allows dynamic parsing of the mib and explicit specification of mib file independent of enviroment variables. called with no args acts like initMib, loading MIBs indicated by environment variables (see Net-SNMP mib_api docs). passing non-zero second arg forces previous mib to be freed and replaced (Note: second arg not working since freeing previous Mib is more involved than before).


calls library init_mib function if Mib not already loaded - does nothing if Mib already loaded. will parse directories and load modules according to environment variables described in Net-SNMP documentations. (see man mib_api, MIBDIRS, MIBS, MIBFILE(S), etc.)


calls library add_mibdir for each directory supplied. will cause directory(s) to be added to internal list and made available for searching in subsequent loadModules calls


calls library read_mib function. The file(s) supplied will be read and all Mib module definitions contained therein will be added to internal mib tree structure


calls library read_module function. The module(s) supplied will be searched for in the current mibdirs and and added to internal mib tree structure. Passing special <mod>, 'ALL', will cause all known modules to be loaded.


*Not Implemented*


will convert a text obj tag to an OID and vice-versa. Any iid suffix is retained numerically. Default behaviour when converting a numeric OID to text form is to return leaf identifier only (e.g.,'sysDescr') but when $SNMP::use_long_names is non-zero or a non-zero second arg is supplied it will return a longer textual identifier. An optional third argument of non-zero will cause the module name to be prepended to the text name (e.g. 'SNMPv2-MIB::sysDescr'). When converting a text obj, the $SNMP::best_guess option is used. If no Mib is loaded when called and $SNMP::auto_init_mib is enabled then the Mib will be loaded. Will return 'undef' upon failure.


return SNMP data type for given textual identifier OBJECTID, OCTETSTR, INTEGER, NETADDR, IPADDR, COUNTER GAUGE, TIMETICKS, OPAQUE, or undef


converts integer value to enumertion tag defined in Mib or converts tag to integer depending on input. the function will return the corresponding integer value *or* tag for a given MIB attribute and value. The function will sense which direction to perform the conversion. Various arg formats are supported

$val = SNMP::mapEnum($varbind);

where $varbind is SNMP::Varbind or equiv. note: $varbind will be updated

$val = SNMP::mapEnum('ipForwarding', 'forwarding');
$val = SNMP::mapEnum('ipForwarding', 1);

Exported SNMP utility functions

Note: utility functions do not support async operation yet.


takes args of SNMP::Session::new followed by those of SNMP::Session::get


takes args of SNMP::Session::new followed by those of SNMP::Session::getnext


takes args of SNMP::Session::new followed by those of SNMP::Session::set


takes args of SNMP::TrapSession::new followed by those of SNMP::TrapSession::trap

Trouble Shooting

If problems occur there are number areas to look at to narrow down the possibilities.

The first step should be to test the Net-SNMP installation independently from the Perl5 SNMP interface.

Try running the apps from the Net-SNMP distribution.

Make sure your agent (snmpd) is running and properly configured with read-write access for the community you are using.

Ensure that your MIBs are installed and enviroment variables are set appropriately (see man mib_api)

Be sure to remove old net-snmp installations and ensure headers and libraries from old CMU installations are not being used by mistake.

If the problem occurs during compilation/linking check that the snmp library being linked is actually the Net-SNMP library (there have been name conflicts with existing snmp libs).

Also check that the header files are correct and up to date.

Sometimes compiling the Net-SNMP library with 'position-independent-code' enabled is required (HPUX specifically).

If you cannot resolve the problem you can post to comp.lang.perl.modules or

please give sufficient information to analyze the problem (OS type, versions for OS/Perl/Net-SNMP/compiler, complete error output, etc.)


Many thanks to all those who supplied patches, suggestions and feedback.

 Joe Marzot (the original author)
 Wes Hardaker and the net-snmp-coders
 Dave Perkins
 Marcel Wiget
 David Blackburn
 John Stofell
 Gary Hayward
 Claire Harrison
 Achim Bohnet
 Doug Kingston
 Jacques Vidrine
 Carl Jacobsen
 Wayne Marquette
 Scott Schumate
 Michael Slifcak
 Srivathsan Srinivasagopalan
 Bill Fenner
 Jef Peeraer
 Daniel Hagerty
 Karl "Rat" Schilke and Electric Lightwave, Inc.
 Perl5 Porters
 Alex Burger

Apologies to any/all who's patch/feature/request was not mentioned or included - most likely it was lost when paying work intruded on my fun. Please try again if you do not see a desired feature. This may actually turn out to be a decent package with such excellent help and the fact that I have more time to work on it than in the past.


bugs, comments, questions to


     Copyright (c) 1995-2000 G. S. Marzot. All rights reserved.
     This program is free software; you can redistribute it and/or
     modify it under the same terms as Perl itself.

     Copyright (c) 2001-2002 Networks Associates Technology, Inc.  All
     Rights Reserved.  This program is free software; you can
     redistribute it and/or modify it under the same terms as Perl