Skip to content

Latest commit

 

History

History
337 lines (265 loc) · 10.2 KB

README.md

File metadata and controls

337 lines (265 loc) · 10.2 KB
Raw

gevent_snmp

Asynchronous SNMP via gevent.

Introduction

Main characteristics as bullet point list:

  • Gives you an 'AsyncSession' object, which works inside a gevent greenlet.
  • Very speedy, ~8000 GET requests per second.
  • It uses the libnetsnmp.so directly via cython (Cython>=0.21 is required).
  • It requires a patched libnetsnmp.so. See implementation notes.
  • Uses the "high level" API of libnetsnmp => Easily use the features provided by libnetsnmp.
  • Tested with gevent 1.1
  • Tested with libnetsnmp 5.7.X

API

The API does no oid translation for you. An oid must be given as a tuple of integers. For example system.sysDescr.0 must be convertet to (1, 3, 6, 1, 2, 1, 1, 1, 0)

The output is always a dictionary containing the varbind list from the response pdu. This varbind list is converted to a dictionary where the keys are the oids (as tuples) and the values the corresponding value from the response pdu.

Create A Session

from async_session import AsyncSession

config = {
    'peername': '127.0.0.1',
    'version': '2c',
    'community': 'public',
    'retries': 5,
    'timeout': 3
}

session = AsyncSession(config)
session.open_session()

oid_str_to_tuple

Converts a oid string of digits into a tuple of integers.

Example:

ret = async_session.oid_str_to_tuple("1.3.6.1.2.1.1.1.0")
assert ret == (1, 3, 6, 1, 2, 1, 1, 1, 0)

oid_tuple_to_str

Converts a tuple of integers to a string with digits and dots.

Example:

ret = async_session.oid_tuple_to_str((1, 3, 6, 1, 2, 1, 1, 1, 0))
assert ret == "1.3.6.1.2.1.1.1.0"

is_in_subtree

Checks if an oid is within another one. First parameter is the root. Second one is checked if its a subtree of the root. Both parameters must be tuple of integers.

ret = async_session.is_in_subtree(
    (1, 3, 6, 1, 4, 1, 4491, 2, 1, 20, 1, 4, 1, 4),
    (1, 3, 6, 1, 4, 1, 4491, 2, 1, 20, 1, 4, 1, 4, 3165, 50798601)
)

assert ret == True

ret = async_session.is_in_subtree(
    (1, 3, 6, 1, 4, 1, 4491, 2, 1, 20, 1, 4, 1, 4),
    (1, 3, 6, 1, 4, 1, 4491, 2, 1, 20, 1, 4, 1, 5, 1)
)

assert ret == False

SNMP-GET

  • Input: List of oid tuples.
  • Output: Dictionary

Example:

oids = [
    (1, 3, 6, 1, 2, 1, 1, 1, 0),
    (1, 3, 6, 1, 2, 1, 1, 3, 0),
]

result = session.get(oids)

print 'system description', result[(1, 3, 6, 1, 2, 1, 1, 1, 0)]
print 'system uptime', result[(1, 3, 6, 1, 2, 1, 1, 3, 0)]

SNMP-GETNEXT

  • Input: A single oid tuple
  • Output Dictionary with the next oid

Example:

oid = (1, 3, 6, 1, 2, 1, 1, 1, 0)
result = session.get_next(oid)
if result:
    next_oid, next_value = result.items()[0]
    print 'next oid is', next_oid
    print 'value for this id', next_value

SNMP-GETBULK

  • Input: List of oid tuples, non repeaters, max repetitions
  • Output: Dictionary

Example:

oids = [
    (1, 3, 6, 1, 2, 1, 1, 1),
    (1, 3, 6, 1, 2, 1, 1, 3),
    (1, 3, 6, 1, 2, 1, 2, 2, 1, 2)
]

# Get
# - system description
# - system uptime
# - The first 5 interface names
result = session.get_bulk(oids, nonrepeaters=2, maxrepetitions=5)

print 'system description', result.pop((1, 3, 6, 1, 2, 1, 1, 1, 0))
print 'system uptime', result.pop((1, 3, 6, 1, 2, 1, 1, 3, 0))

for oid, ifname in result.items():
    print 'oid', oid
    print 'interface name', ifname

WALK

  • Input: A single oid tuple
  • Output Dictionary having all oids which are childs of the input oid

This walk uses only snmp-getnext to traverse the tree.

Example:

root_id = (1, 3, 6, 1, 2, 1, 1)
result = session.walk(root_id)
for oid in sorted(result):
    print 'oid', oid, result[oid]

WALK with GETBULK

  • Input: A single oid tuple, how much oids to retrieve in a single getbulk
  • Output Dictionary having all (full) oids which are childs of the input oid

The main difference to walk is the usage of the snmp-getbulk operation to traverse the tree.

Example:

root_id = (1, 3, 6, 1, 2, 1, 1)
result = session.walk_with_get_bulk(root_id, maxrepetitions=10)
for oid in sorted(result):
    print 'oid', oid, result[oid]

SET

  • Input: Dictionary with oid tuples as keys. The values are tuples again, where the first element is the value to set as a string and the second parameter a type specifiction. The type specification is used to encode the value correctly inside the pdu. Have a look here for the possible type specifications. https://linux.die.net/man/1/snmpset The library takes care to encode the value string according to the type specification.

Example:

to_set = {
    # Set a new location.
    (1, 3, 6, 1, 2, 1, 1, 1, 0): ('new location', 's'),
    
    # Also integers have to be given as strings.
    # The following disable IP forwarding on the first interface.
    (1, 3, 6, 1, 2, 1, 4, 1, 0): ("2", "i")
}

result = session.set_oids(to_set)

Error Handling

The APIs mentioned above may raise the following exceptions.

  • SNMPTimeoutError
  • SNMPResponseError

SNMPTimeoutError

This exception is raised if there is no SNMP-Response within the configured time constraints.

SNMPResponseError

This exception is raised if something with the response is wrong. It has the following attributes:

  • code: Which is the error status from the response PDU
  • index: Which is the error index from the response PDU
  • message: String representation of the error
try:
    session.set_oids(to_set)
except SNMPResponseError as error:
    print error.code
    print error.index
    print error.message

Flags

The following methods have an optional py_flags parameter to control how the response is parsed.

  • walk
  • walk_with_get_bulk
  • get
  • get_next
  • get_bulk
  • set

This parameter is a python dictionary which may contains the following flags:

get_var_type

Each entry in the varbind list of a SNMP response contains type and value. Per default the API takes automatically care to convert the value into the corresponding python object. However it is also possible to get the type for each entry. Is get_var_type given and set to True the value for the returned dictionary will be a tuple. Where the first element is the type and the second element the value as a python object.

For example:

for oid, (asn_type, asn_value) for session.walk(oid, py_flags={'get_var_type': True}).items():
    print oid, asn_type, asn_value

get_end_of_mib

If given and set to True each single varbind of type SNMP_ENDOFMIBVIEW will be encode to the special object async_session.END_OF_MIB. If this flag is not set 'end of mib view' is converted to None.

get_no_such_object

If given and set to True each single varbind of type SNMP_NOSUCHOBJECT: will be encode to the special object async_session.NO_SUCH_OBJECT. If this flag is not set 'no such object' is converted to None.

get_no_such_instance

If given and set to True each single varbind of type SNMP_NOSUCHINSTANCE will be encode to the special object async_session.NO_SUCH_INSTANCE. If this flag is not set 'no such instance' is converted to None.

as_ordered_dict

If given and set to True a collections.OrderedDict is returned. It preserves the order of the incoming varbinds.

as_list

If given and set to True the result is return as a list of (oid, value) tuples.

get_netsnmp_string

If given and set to True each result value is converted to a string. This string is generated by the NET-SNMP function family print_value. These functions take the definition of an OID from its corresponding MIB to format the value accordingly. NET-SNMP offers some options to control the format behaviour. See toggle_netsnmp_format_options. The result of the SNMP-API changes then to a tuple where the raw value is at position one and the generated string at position two.

Note: The MIB's must be loaded, otherwise the format does not work. To do so call the following at process startup:

async_session.init_snmplib()

Clone

Use this call to clone an existing session.

  • Input: Config options for the new session. Options not mentioned here are cloned from the existing session. The overriding options must be given as keyword arguments.
  • Output: Contextmanager which delivers the cloned session. For this clone open_session has been already called.

Example:

with sess.clone_session(community='private') as priv_sess:
    print priv_sess.set_oids(oids)

toggle_netsnmp_format_options

Control the format of the values if the flag get_netsnmp_string=True is used. Since the NET-SNMP functions for formatting are used, the corresponding format options are also supported. To get a list of all supported options see:

For example to enable 'quick view' and 'numeric oids use:

async_session.toggle_netsnmp_format_options('nq')

Keep in mind:

This function toggles the option. For example:

# Enables 'quick view'
async_session.toggle_netsnmp_format_options('q')

# Disables 'quick view'
async_session.toggle_netsnmp_format_options('q')

Furthermore these options are global for the entire process. So they affect all AsyncSession() at the same time. Hence the recommendation is to set them once at process startup.

Implementation Notes

Why the patch ? gevent_snmp works by replacing the calls to 'select()' inside libnetsnmp. So if libnetsnmp would call 'select()', gevent.socket.wait_read is called insted. Per default libnetsnmp does not allow to replace/override the calls to select. The patch to libnetsnmp just adds another function where the select could be replaced via a callback. The advantage of this patch is, that we can still use the synchronous 'high level' API of libnetsnmp. This synchronous 'high levell' is way more easier to use than the asynchronous 'low level' API. For example we get retry and timeout handling for free, because the synchronous 'high level' API implements them. Whereas the asynchronous 'low level' API does not support retry/timeout.

Have a look at netsnmp_patch.diff, its fairly easy and not intrusive at all.