olsrd

#!/bin/sh
# weird shebang? See below: "interpreter selection"

"""true"
: <<=cut

=head1 NAME

olsrd - Monitor the state of an OLSR-based routing network


=head1 APPLICABLE SYSTEMS

Information is parsed from the output of "txtinfo" plugin for olsrd.


=head1 CONFIGURATION

Environment variables:

    * OLSRD_HOST: name or IP of the host running the txtinfo plugin (default: localhost)
    * OLSRD_TXTINFO_PORT: the port that the txtinfo plugin is listening to (default: 2006)
    * OLSRD_BIN_PATH: name of the olsrd binary (only used for 'autoconf', default: /usr/sbin/olsrd)
    * MICROPYTHON_HEAP: adjust this parameter for micropython if your olsr network contains
      more than a few thousand nodes (default: 512k)

=head1 USAGE

Collect basic information about the neighbours of an OLSR node:

    * link quality
    * neighbour link quality
    * number of nodes reachable behind each neighbour
    * ping times of direct neighbours

This plugin works with the following python interpreters:

    * Python 2
    * Python 3
    * micropython (e.g. OpenWrt)


=head1 VERSION

  0.4


=head1 AUTHOR

Lars Kruse <devel@sumpfralle.de>


=head1 LICENSE

GPLv3 or above


=head1 MAGIC MARKERS

  #%# family=auto
  #%# capabilities=autoconf

=cut


# ****************** Interpreter Selection ***************
# This unbelievable dirty hack allows to find a suitable python interpreter.
# This is specifically useful for OpenWRT where typically only micropython is available.
#
# Additionally we need to run micropython with additional startup options.
# This is necessary due to our demand for more than 128k heap (this default is sufficient for only
# 400 olsr nodes).
#
# This "execution hack" works as follows:
#   * the script is executed by busybox ash or another shell
#   * the above line (three quotes before and one quote after 'true') evaluates differently for
#     shell and python:
#       * shell: run "true" (i.e. nothing happens)
#       * python: ignore everything up to the next three consecutive quotes
# Thus we may place shell code here that will take care for selecting an interpreter.

# prefer micropython if it is available - otherwise fall back to any python (2 or 3)
MICROPYTHON_BIN=$(which micropython || true)
if [ -n "$MICROPYTHON_BIN" ]; then
    "$MICROPYTHON_BIN" -X "heapsize=${MICROPYTHON_HEAP:-512k}" "$0" "$@"
else
    python "$0" "$@"
fi
exit $?

# For shell: ignore everything starting from here until the last line of this file.
# This is necessary for syntax checkers that try to complain about invalid shell syntax below.
true <<EOF
"""


import os
import os.path
import socket
import sys


plugin_version = "0.4"

LQ_GRAPH_CONFIG = """
graph_title     {title}
graph_vlabel    Link Quality (-) / Neighbour Link Quality (+)
graph_category  network
graph_info      OLSR estimates the quality of a connection by the ratio of successfully received \
(link quality) and transmitted (neighbour link quality) hello packets.
"""

LQ_VALUES_CONFIG = """
nlq{suffix}.label none
nlq{suffix}.type GAUGE
nlq{suffix}.graph no
nlq{suffix}.draw {draw_type}
nlq{suffix}.min 0
lq{suffix}.label {label}
lq{suffix}.type GAUGE
lq{suffix}.draw {draw_type}
lq{suffix}.negative nlq{suffix}
lq{suffix}.min 0
"""

NEIGHBOUR_COUNT_CONFIG = """
graph_title     Reachable nodes via neighbours
graph_vlabel    Number of Nodes
graph_category  network
graph_info      Count the number of locally known routes passing through each direct neighbour. \
This number is a good approximation for the number of mesh nodes reachable via this specific \
neighbour. MIDs (alternative addresses of an OLSR node) and HNAs (host network announcements) are \
ignored.
"""

NEIGHBOUR_COUNT_VALUE = """
neighbour_{host_fieldname}.label {host}
neighbour_{host_fieldname}.type GAUGE
neighbour_{host_fieldname}.draw {draw_type}
neighbour_{host_fieldname}.min 0
"""

NEIGHBOUR_PING_CONFIG = """
graph_title     {title}
graph_vlabel    roundtrip time (ms)
graph_category  network
graph_info      This graph shows ping RTT statistics.
graph_args      --base 1000 --lower-limit 0
graph_scale     no
"""

NEIGHBOUR_PING_VALUE = """neighbour_{host_fieldname}.label {host}"""

# micropython (as of 2015) does not contain "os.linesep"
LINESEP = getattr(os, "linesep", "\n")


def get_clean_fieldname(name):
    chars = []
    for index, char in enumerate(name):
        if ("a" <= char.lower() <= "z") or ((index == 0) or ("0" <= char <= "9")):
            chars.append(char)
        else:
            chars.append("_")
    return "".join(chars)


def query_olsrd_txtservice(section=""):
    host = os.getenv("OLSRD_HOST", "localhost")
    port = os.getenv("OLSRD_TXTINFO_PORT", "2006")
    conn = socket.create_connection((host, port), 1.0)
    try:
        # Python3
        request = bytes("/%s" % section, "ascii")
    except TypeError:
        # Python2
        request = bytes("/%s" % section)
    conn.sendall(request)
    fconn = conn.makefile()
    in_header = True
    in_body_count = 0
    for line in fconn.readlines():
        if in_header:
            if not line.strip():
                # the empty line marks the end of the header
                in_header = False
            # ignore header lines (nothing to be done)
        else:
            # skip the first two body lines - they are table headers
            if in_body_count >= 2:
                line = line.strip()
                if line:
                    yield line
            in_body_count += 1
    fconn.close()
    conn.close()


def get_address_device_mapping():
    mapping = {}
    for line in query_olsrd_txtservice("mid"):
        # example line content:
        #    192.168.2.171   192.168.22.171;192.168.12.171
        # since olsr v0.9.5:
        #    192.168.2.171   192.168.22.171    192.168.12.171
        device_id, mids = line.split(None, 1)
        for mid in mids.replace(";", " ").split():
            mapping[mid] = device_id
    return mapping


def count_routes_by_neighbour(address_mapping, ignore_list):
    node_count = {}
    for line in query_olsrd_txtservice("rou"):
        # example line content:
        #    192.168.1.79/32 192.168.12.38   4       4.008   wlan0
        tokens = line.split()
        target = tokens[0]
        via = tokens[1]
        # we care only about single-host routes
        if target.endswith("/32"):
            if target[:-3] in address_mapping:
                # we ignore MIDs - we want only real nodes
                continue
            if target in ignore_list:
                continue
            # replace the neighbour's IP with its main IP (if it is an MID)
            via = address_mapping.get(via, via)
            # increase the counter
            node_count[via] = node_count.get(via, 0) + 1
    return node_count


def get_olsr_links():
    mid_mapping = get_address_device_mapping()
    hna_list = [line.split()[0] for line in query_olsrd_txtservice("hna")]
    route_count = count_routes_by_neighbour(mid_mapping, hna_list)
    result = []
    for line in query_olsrd_txtservice("lin"):
        tokens = line.split()
        # the "cost" may be infinite
        if tokens[-1] == "INFINITE":
            # "inf" is the python keyword for "maximum float number"
            tokens[-1] = "inf"
        link = {}
        link["local"] = tokens.pop(0)
        remote = tokens.pop(0)
        # replace the neighbour's IP with its main IP (if it is an MID)
        link["remote"] = mid_mapping.get(remote, remote)
        for key in ("hysterese", "lq", "nlq", "cost"):
            link[key] = float(tokens.pop(0))
        # add the route count
        link["route_count"] = route_count.get(link["remote"], 0)
        result.append(link)
    result.sort(key=lambda link: link["remote"])
    return result


def _read_file(filename):
    try:
        return open(filename, "r").read().split(LINESEP)
    except OSError:
        return []


def get_ping_times(hosts):
    tempfile = "/tmp/munin-olsrd-{pid}.tmp".format(pid=os.getpid())
    command = ('for host in {hosts}; do echo -n "$host "; '
               'ping -c 1 -w 1 "$host" | grep /avg/ || echo; done >{tempfile}'
               .format(hosts=" ".join(hosts), tempfile=tempfile))
    # micropython supports only "os.system" (as of 2015) - thus we need to stick with it for
    # OpenWrt.
    returncode = os.system(command)
    if returncode != 0:
        return {}
    lines = _read_file(tempfile)
    os.unlink(tempfile)
    # example output for one host:
    #   192.168.2.41 round-trip min/avg/max = 4.226/4.226/4.226 ms
    result = {}
    for line in lines:
        tokens = line.split(None)
        if len(tokens) > 1:
            host = tokens[0]
            avg_ping = tokens[-2].split("/")[1]
            result[host] = float(avg_ping)
    return result


def do_config():
    links = list(get_olsr_links())

    # link quality with regard to neighbours
    print("multigraph olsr_link_quality")
    print(LQ_GRAPH_CONFIG.format(title="OLSR Link Quality"))
    for link in links:
        print(LQ_VALUES_CONFIG.format(
            label=link["remote"],
            suffix="_{host}".format(host=get_clean_fieldname(link["remote"])),
            draw_type="AREASTACK"))
    for link in links:
        print("multigraph olsr_link_quality.host_{remote}"
              .format(remote=get_clean_fieldname(link["remote"])))
        title = "Link Quality towards {host}".format(host=link["remote"])
        print(LQ_GRAPH_CONFIG.format(title=title))
        print(LQ_VALUES_CONFIG.format(label="Link Quality", suffix="", draw_type="AREA"))

    # link count ("number of nodes behind each neighbour")
    print("multigraph olsr_neighbour_link_count")
    print(NEIGHBOUR_COUNT_CONFIG)
    for link in links:
        print(NEIGHBOUR_COUNT_VALUE
              .format(host=link["remote"], host_fieldname=get_clean_fieldname(link["remote"]),
                      draw_type="AREASTACK"))

    # neighbour ping
    print("multigraph olsr_neighbour_ping")
    print(NEIGHBOUR_PING_CONFIG.format(title="Ping time of neighbours"))
    for link in links:
        print(NEIGHBOUR_PING_VALUE
              .format(host=link["remote"], host_fieldname=get_clean_fieldname(link["remote"])))
    # neighbour pings - single subgraphs
    for link in links:
        remote = get_clean_fieldname(link["remote"])
        print("multigraph olsr_neighbour_ping.host_{remote}".format(remote=remote))
        title = "Ping time of {remote}".format(remote=remote)
        print(NEIGHBOUR_PING_CONFIG.format(title=title))
        print(NEIGHBOUR_PING_VALUE.format(host=link["remote"], host_fieldname=remote))


def do_fetch():
    # output values
    links = list(get_olsr_links())

    # overview graph for the link quality (ETX) of all neighbours
    print("multigraph olsr_link_quality")
    for link in links:
        print("lq_{remote}.value {lq:f}".format(lq=link["lq"],
                                                remote=get_clean_fieldname(link["remote"])))
        print("nlq_{remote}.value {nlq:f}".format(nlq=link["nlq"],
                                                  remote=get_clean_fieldname(link["remote"])))
    # detailed ETX graph for each single neighbour link
    for link in links:
        print("multigraph olsr_link_quality.host_{remote}"
              .format(remote=get_clean_fieldname(link["remote"])))
        print("lq.value {lq:f}".format(lq=link["lq"]))
        print("nlq.value {nlq:f}".format(nlq=link["nlq"]))

    # count the links/nodes behind each neighbour node
    print("multigraph olsr_neighbour_link_count")
    for link in links:
        print("neighbour_{host_fieldname}.value {value}"
              .format(value=link["route_count"],
                      host_fieldname=get_clean_fieldname(link["remote"])))

    # overview of ping roundtrip times
    print("multigraph olsr_neighbour_ping")
    ping_times = get_ping_times([link["remote"] for link in links])
    for link in links:
        ping_time = ping_times.get(link["remote"], None)
        value = "{:.4f}".format(ping_time) if ping_time is not None else "U"
        print("neighbour_{remote}.value {value}"
              .format(value=value, remote=get_clean_fieldname(link["remote"])))
    # single detailed graphs for the ping time of each link
    for link in links:
        ping_time = ping_times.get(link["remote"], None)
        value = "{:.4f}".format(ping_time) if ping_time is not None else "U"
        remote = get_clean_fieldname(link["remote"])
        print("multigraph olsr_neighbour_ping.host_{remote}".format(remote=remote))
        print("neighbour_{remote}.value {value}".format(remote=remote, value=value))


if __name__ == "__main__":
    # parse arguments
    if len(sys.argv) > 1:
        if sys.argv[1] == "config":
            do_config()
            if os.getenv("MUNIN_CAP_DIRTYCONFIG") == "1":
                do_fetch()
            sys.exit(0)
        elif sys.argv[1] == "autoconf":
            if os.path.exists(os.getenv('OLSRD_BIN_PATH', '/usr/sbin/olsrd')):
                print('yes')
            else:
                print('no')
            sys.exit(0)
        elif sys.argv[1] == "version":
            print('olsrd Munin plugin, version %s' % plugin_version)
            sys.exit(0)
        elif sys.argv[1] == "":
            # ignore
            pass
        else:
            # unknown argument
            sys.stderr.write("Unknown argument{eol}".format(eol=LINESEP))
            sys.exit(1)

    do_fetch()

# final marker for shell / python hybrid script (see "Interpreter Selection")
EOF = True
EOF