path: root/tools/collectors/collectd/collectd_bucky.py

# Copyright 2014-2018 TRBS, Spirent Communications
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may not
# use this file except in compliance with the License. You may obtain a copy of
# the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations under
# the License.

# This file is a modified version of scripts present in bucky software
# details of bucky can be found at https://github.com/trbs/bucky

"""
This module receives the samples from collectd, processes it and
enqueues it in a format suitable for easy processing.
It also handles secure communication with collectd.
"""
import copy
import hmac
import logging
import multiprocessing
import os
import socket
import struct
import sys
from hashlib import sha1, sha256

from Crypto.Cipher import AES
from conf import settings

logging.basicConfig()
LOG = logging.getLogger(__name__)


class CollectdError(Exception):
    """
    Custom error class.
    """
    def __init__(self, mesg):
        super(CollectdError, self).__init__(mesg)
        self.mesg = mesg

    def __str__(self):
        return self.mesg


class ConnectError(CollectdError):
    """
    Custom connect error
    """
    pass


class ConfigError(CollectdError):
    """
    Custom config error
    """
    pass


class ProtocolError(CollectdError):
    """
    Custom protocol error
    """
    pass


class UDPServer(multiprocessing.Process):
    """
    Actual UDP server receiving collectd samples over network
    """
    def __init__(self, ip, port):
        super(UDPServer, self).__init__()
        self.daemon = True
        addrinfo = socket.getaddrinfo(ip, port,
                                      socket.AF_UNSPEC, socket.SOCK_DGRAM)
        afamily, _, _, _, addr = addrinfo[0]
        ip, port = addr[:2]
        self.ip_addr = ip
        self.port = port
        self.sock = socket.socket(afamily, socket.SOCK_DGRAM)
        self.sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
        try:
            self.sock.bind((ip, port))
            LOG.info("Bound socket socket %s:%s", ip, port)
        except socket.error:
            LOG.exception("Error binding socket %s:%s.", ip, port)
            sys.exit(1)

        self.sock_recvfrom = self.sock.recvfrom

    def run(self):
        """
        Start receiving messages
        """
        recvfrom = self.sock_recvfrom
        while True:
            try:
                data, addr = recvfrom(65535)
            except (IOError, KeyboardInterrupt):
                continue
            addr = addr[:2]  # for compatibility with longer ipv6 tuples
            if data == b'EXIT':
                break
            if not self.handle(data, addr):
                break
        try:
            self.pre_shutdown()
        except SystemExit:
            LOG.exception("Failed pre_shutdown method for %s",
                          self.__class__.__name__)

    def handle(self, data, addr):
        """
        Handle the message.
        """
        raise NotImplementedError()

    def pre_shutdown(self):
        """ Pre shutdown hook """
        pass

    def close(self):
        """
        Close the communication
        """
        self.send('EXIT')

    def send(self, data):
        """
        Send over the network
        """
        sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
        if not isinstance(data, bytes):
            data = data.encode()
        sock.sendto(data, 0, (self.ip_addr, self.port))


class CPUConverter(object):
    """
    Converter for CPU samples fom collectd.
    """
    PRIORITY = -1

    def __call__(self, sample):
        return ["cpu", sample["plugin_instance"], sample["type_instance"]]


class InterfaceConverter(object):
    """
    Converter for Interface samples from collectd
    """
    PRIORITY = -1

    def __call__(self, sample):
        parts = []
        parts.append("interface")
        if sample.get("plugin_instance", ""):
            parts.append(sample["plugin_instance"].strip())
        stypei = sample.get("type_instance", "").strip()
        if stypei:
            parts.append(stypei)
        stype = sample.get("type").strip()
        if stype:
            parts.append(stype)
        vname = sample.get("value_name").strip()
        if vname:
            parts.append(vname)
        return parts


class MemoryConverter(object):
    """
    Converter for Memory samples from collectd
    """
    PRIORITY = -1

    def __call__(self, sample):
        return ["memory", sample["type_instance"]]


class DefaultConverter(object):
    """
    Default converter for samples from collectd
    """
    PRIORITY = -1

    def __call__(self, sample):
        parts = []
        parts.append(sample["plugin"].strip())
        if sample.get("plugin_instance"):
            parts.append(sample["plugin_instance"].strip())
        stype = sample.get("type", "").strip()
        if stype and stype != "value":
            parts.append(stype)
        stypei = sample.get("type_instance", "").strip()
        if stypei:
            parts.append(stypei)
        vname = sample.get("value_name").strip()
        if vname and vname != "value":
            parts.append(vname)
        return parts


DEFAULT_CONVERTERS = {
    "cpu": CPUConverter(),
    "interface": InterfaceConverter(),
    "memory": MemoryConverter(),
    "_default": DefaultConverter(),
}


class CollectDTypes(object):
    """
    Class to handle the sample types. The types.db that comes
    with collectd, usually, defines the various types.
    """
    def __init__(self, types_dbs=None):
        if types_dbs is None:
            types_dbs = []
        dirs = ["/opt/collectd/share/collectd/types.db",
                "/usr/local/share/collectd/types.db"]
        self.types = {}
        self.type_ranges = {}
        if not types_dbs:
            types_dbs = [tdb for tdb in dirs if os.path.exists(tdb)]
            if not types_dbs:
                raise ConfigError("Unable to locate types.db")
        self.types_dbs = types_dbs
        self._load_types()

    def get(self, name):
        """
        Get the name of the type
        """
        t_name = self.types.get(name)
        if t_name is None:
            raise ProtocolError("Invalid type name: %s" % name)
        return t_name

    def _load_types(self):
        """
        Load all the types from types_db
        """
        for types_db in self.types_dbs:
            with open(types_db) as handle:
                for line in handle:
                    if line.lstrip()[:1] == "#":
                        continue
                    if not line.strip():
                        continue
                    self._add_type_line(line)
            LOG.info("Loaded collectd types from %s", types_db)

    def _add_type_line(self, line):
        """
        Add types information
        """
        types = {
            "COUNTER": 0,
            "GAUGE": 1,
            "DERIVE": 2,
            "ABSOLUTE": 3
        }
        name, spec = line.split(None, 1)
        self.types[name] = []
        self.type_ranges[name] = {}
        vals = spec.split(", ")
        for val in vals:
            vname, vtype, minv, maxv = val.strip().split(":")
            vtype = types.get(vtype)
            if vtype is None:
                raise ValueError("Invalid value type: %s" % vtype)
            minv = None if minv == "U" else float(minv)
            maxv = None if maxv == "U" else float(maxv)
            self.types[name].append((vname, vtype))
            self.type_ranges[name][vname] = (minv, maxv)


class CollectDParser(object):
    """
    Parser class: Implements the sample parsing operations.
    The types definition defines the parsing process.
    """
    def __init__(self, types_dbs=None, counter_eq_derive=False):
        if types_dbs is None:
            types_dbs = []
        self.types = CollectDTypes(types_dbs=types_dbs)
        self.counter_eq_derive = counter_eq_derive

    def parse(self, data):
        """
        Parse individual samples
        """
        for sample in self.parse_samples(data):
            yield sample

    def parse_samples(self, data):
        """
        Extract all the samples from the message.
        """
        types = {
            0x0000: self._parse_string("host"),
            0x0001: self._parse_time("time"),
            0x0008: self._parse_time_hires("time"),
            0x0002: self._parse_string("plugin"),
            0x0003: self._parse_string("plugin_instance"),
            0x0004: self._parse_string("type"),
            0x0005: self._parse_string("type_instance"),
            0x0006: None,  # handle specially
            0x0007: self._parse_time("interval"),
            0x0009: self._parse_time_hires("interval")
        }
        sample = {}
        for (ptype, pdata) in self.parse_data(data):
            if ptype not in types:
                LOG.debug("Ignoring part type: 0x%02x", ptype)
                continue
            if ptype != 0x0006:
                types[ptype](sample, pdata)
                continue
            for vname, vtype, val in self.parse_values(sample["type"], pdata):
                sample["value_name"] = vname
                sample["value_type"] = vtype
                sample["value"] = val
                yield copy.deepcopy(sample)

    @staticmethod
    def parse_data(data):
        """
        Parse the message
        """
        types = set([
            0x0000, 0x0001, 0x0002, 0x0003, 0x0004,
            0x0005, 0x0006, 0x0007, 0x0008, 0x0009,
            0x0100, 0x0101, 0x0200, 0x0210
        ])
        while data:
            if len(data) < 4:
                raise ProtocolError("Truncated header.")
            (part_type, part_len) = struct.unpack("!HH", data[:4])
            data = data[4:]
            if part_type not in types:
                raise ProtocolError("Invalid part type: 0x%02x" % part_type)
            part_len -= 4  # includes four header bytes we just parsed
            if len(data) < part_len:
                raise ProtocolError("Truncated value.")
            part_data, data = data[:part_len], data[part_len:]
            yield (part_type, part_data)

    def parse_values(self, stype, data):
        """
        Parse the value of a particular type
        """
        types = {0: "!Q", 1: "<d", 2: "!q", 3: "!Q"}
        (nvals,) = struct.unpack("!H", data[:2])
        data = data[2:]
        if len(data) != 9 * nvals:
            raise ProtocolError("Invalid value structure length.")
        vtypes = self.types.get(stype)
        if nvals != len(vtypes):
            raise ProtocolError("Values different than types.db info.")
        for i in range(nvals):
            vtype = data[i]
            if vtype != vtypes[i][1]:
                if self.counter_eq_derive and \
                   (vtype, vtypes[i][1]) in ((0, 2), (2, 0)):
                    # if counter vs derive don't break, assume server is right
                    LOG.debug("Type mismatch (counter/derive) for %s/%s",
                              stype, vtypes[i][0])
                else:
                    raise ProtocolError("Type mismatch with types.db")
        data = data[nvals:]
        for i in range(nvals):
            vdata, data = data[:8], data[8:]
            (val,) = struct.unpack(types[vtypes[i][1]], vdata)
            yield vtypes[i][0], vtypes[i][1], val

    @staticmethod
    def _parse_string(name):
        """
        Parse string value
        """
        def _parser(sample, data):
            """
            Actual string parser
            """
            data = data.decode()
            if data[-1] != '\0':
                raise ProtocolError("Invalid string detected.")
            sample[name] = data[:-1]
        return _parser

    @staticmethod
    def _parse_time(name):
        """
        Parse time value
        """
        def _parser(sample, data):
            """
            Actual time parser
            """
            if len(data) != 8:
                raise ProtocolError("Invalid time data length.")
            (val,) = struct.unpack("!Q", data)
            sample[name] = float(val)
        return _parser

    @staticmethod
    def _parse_time_hires(name):
        """
        Parse time hires value
        """
        def _parser(sample, data):
            """
            Actual time hires parser
            """
            if len(data) != 8:
                raise ProtocolError("Invalid hires time data length.")
            (val,) = struct.unpack("!Q", data)
            sample[name] = val * (2 ** -30)
        return _parser


class CollectDCrypto(object):
    """
    Handle the sercured communications with collectd daemon
    """
    def __init__(self):
        sec_level = settings.getValue('COLLECTD_SECURITY_LEVEL')
        if sec_level in ("sign", "SIGN", "Sign", 1):
            self.sec_level = 1
        elif sec_level in ("encrypt", "ENCRYPT", "Encrypt", 2):
            self.sec_level = 2
        else:
            self.sec_level = 0
        if self.sec_level:
            self.auth_file = settings.getValue('COLLECTD_AUTH_FILE')
            self.auth_db = {}
            if self.auth_file:
                self.load_auth_file()
            if not self.auth_file:
                raise ConfigError("Collectd security level configured but no "
                                  "auth file specified in configuration")
            if not self.auth_db:
                LOG.warning("Collectd security level configured but no "
                            "user/passwd entries loaded from auth file")

    def load_auth_file(self):
        """
        Loading the authentication file.
        """
        try:
            fil = open(self.auth_file)
        except IOError as exc:
            raise ConfigError("Unable to load collectd's auth file: %r" % exc)
        self.auth_db.clear()
        for line in fil:
            line = line.strip()
            if not line or line[0] == "#":
                continue
            user, passwd = line.split(":", 1)
            user = user.strip()
            passwd = passwd.strip()
            if not user or not passwd:
                LOG.warning("Found line with missing user or password")
                continue
            if user in self.auth_db:
                LOG.warning("Found multiple entries for single user")
            self.auth_db[user] = passwd
        fil.close()
        LOG.info("Loaded collectd's auth file from %s", self.auth_file)

    def parse(self, data):
        """
        Parse the non-encrypted message
        """
        if len(data) < 4:
            raise ProtocolError("Truncated header.")
        part_type, part_len = struct.unpack("!HH", data[:4])
        sec_level = {0x0200: 1, 0x0210: 2}.get(part_type, 0)
        if sec_level < self.sec_level:
            raise ProtocolError("Packet has lower security level than allowed")
        if not sec_level:
            return data
        if sec_level == 1 and not self.sec_level:
            return data[part_len:]
        data = data[4:]
        part_len -= 4
        if len(data) < part_len:
            raise ProtocolError("Truncated part payload.")
        if sec_level == 1:
            return self.parse_signed(part_len, data)
        if sec_level == 2:
            return self.parse_encrypted(part_len, data)

    def parse_signed(self, part_len, data):
        """
        Parse the signed message
        """

        if part_len <= 32:
            raise ProtocolError("Truncated signed part.")
        sig, data = data[:32], data[32:]
        uname_len = part_len - 32
        uname = data[:uname_len].decode()
        if uname not in self.auth_db:
            raise ProtocolError("Signed packet, unknown user '%s'" % uname)
        password = self.auth_db[uname].encode()
        sig2 = hmac.new(password, msg=data, digestmod=sha256).digest()
        if not self._hashes_match(sig, sig2):
            raise ProtocolError("Bad signature from user '%s'" % uname)
        data = data[uname_len:]
        return data

    def parse_encrypted(self, part_len, data):
        """
        Parse the encrypted message
        """
        if part_len != len(data):
            raise ProtocolError("Enc pkt size disaggrees with header.")
        if len(data) <= 38:
            raise ProtocolError("Truncated encrypted part.")
        uname_len, data = struct.unpack("!H", data[:2])[0], data[2:]
        if len(data) <= uname_len + 36:
            raise ProtocolError("Truncated encrypted part.")
        uname, data = data[:uname_len].decode(), data[uname_len:]
        if uname not in self.auth_db:
            raise ProtocolError("Couldn't decrypt, unknown user '%s'" % uname)
        ival, data = data[:16], data[16:]
        password = self.auth_db[uname].encode()
        key = sha256(password).digest()
        pad_bytes = 16 - (len(data) % 16)
        data += b'\0' * pad_bytes
        data = AES.new(key, IV=ival, mode=AES.MODE_OFB).decrypt(data)
        data = data[:-pad_bytes]
        tag, data = data[:20], data[20:]
        tag2 = sha1(data).digest()
        if not self._hashes_match(tag, tag2):
            raise ProtocolError("Bad checksum on enc pkt for '%s'" % uname)
        return data

    @staticmethod
    def _hashes_match(val_a, val_b):
        """Constant time comparison of bytes """
        if len(val_a) != len(val_b):
            return False
        diff = 0
        for val_x, val_y in zip(val_a, val_b):
            diff |= val_x ^ val_y
        return not diff


class CollectDConverter(object):
    """
    Handle all conversions.
    Coversion: Convert the sample received from collectd to an
    appropriate format - for easy processing
    """
    def __init__(self):
        self.converters = dict(DEFAULT_CONVERTERS)

    def convert(self, sample):
        """
        Main conversion handling.
        """
        default = self.converters["_default"]
        handler = self.converters.get(sample["plugin"], default)
        try:
            name_parts = handler(sample)
            if name_parts is None:
                return  # treat None as "ignore sample"
            name = '.'.join(name_parts)
        except (AttributeError, IndexError, MemoryError, RuntimeError):
            LOG.exception("Exception in sample handler  %s (%s):",
                          sample["plugin"], handler)
            return
        host = sample.get("host", "")
        return (
            host,
            name,
            sample["value_type"],
            sample["value"],
            int(sample["time"])
        )

    def _add_converter(self, name, inst, source="unknown"):
        """
        Add new converter types
        """
        if name not in self.converters:
            LOG.info("Converter: %s from %s", name, source)
            self.converters[name] = inst
            return
        kpriority = getattr(inst, "PRIORITY", 0)
        ipriority = getattr(self.converters[name], "PRIORITY", 0)
        if kpriority > ipriority:
            LOG.info("Replacing: %s", name)
            LOG.info("Converter: %s from %s", name, source)
            self.converters[name] = inst
            return
        LOG.info("Ignoring: %s (%s) from %s (priority: %s vs %s)",
                 name, inst, source, kpriority, ipriority)


class CollectDHandler(object):
    """Wraps all CollectD parsing functionality in a class"""

    def __init__(self):
        self.crypto = CollectDCrypto()
        collectd_types = []
        collectd_counter_eq_derive = False
        self.parser = CollectDParser(collectd_types,
                                     collectd_counter_eq_derive)
        self.converter = CollectDConverter()
        self.prev_samples = {}
        self.last_sample = None

    def parse(self, data):
        """
        Parse the samples from collectd
        """
        try:
            data = self.crypto.parse(data)
        except ProtocolError as error:
            LOG.error("Protocol error in CollectDCrypto: %s", error)
            return
        try:
            for sample in self.parser.parse(data):
                self.last_sample = sample
                stype = sample["type"]
                vname = sample["value_name"]
                sample = self.converter.convert(sample)
                if sample is None:
                    continue
                host, name, vtype, val, time = sample
                if not name.strip():
                    continue
                val = self.calculate(host, name, vtype, val, time)
                val = self.check_range(stype, vname, val)
                if val is not None:
                    yield host, name, val, time
        except ProtocolError as error:
            LOG.error("Protocol error: %s", error)
            if self.last_sample is not None:
                LOG.info("Last sample: %s", self.last_sample)

    def check_range(self, stype, vname, val):
        """
        Check the value range
        """
        if val is None:
            return
        try:
            vmin, vmax = self.parser.types.type_ranges[stype][vname]
        except KeyError:
            LOG.error("Couldn't find vmin, vmax in CollectDTypes")
            return val
        if vmin is not None and val < vmin:
            LOG.debug("Invalid value %s (<%s) for %s", val, vmin, vname)
            LOG.debug("Last sample: %s", self.last_sample)
            return
        if vmax is not None and val > vmax:
            LOG.debug("Invalid value %s (>%s) for %s", val, vmax, vname)
            LOG.debug("Last sample: %s", self.last_sample)
            return
        return val

    def calculate(self, host, name, vtype, val, time):
        """
        Perform calculations for handlers
        """
        handlers = {
            0: self._calc_counter,  # counter
            1: lambda _host, _name, v, _time: v,  # gauge
            2: self._calc_derive,  # derive
            3: self._calc_absolute  # absolute
        }
        if vtype not in handlers:
            LOG.error("Invalid value type %s for %s", vtype, name)
            LOG.info("Last sample: %s", self.last_sample)
            return
        return handlers[vtype](host, name, val, time)

    def _calc_counter(self, host, name, val, time):
        """
        Calculating counter values
        """
        key = (host, name)
        if key not in self.prev_samples:
            self.prev_samples[key] = (val, time)
            return
        pval, ptime = self.prev_samples[key]
        self.prev_samples[key] = (val, time)
        if time <= ptime:
            LOG.error("Invalid COUNTER update for: %s:%s", key[0], key[1])
            LOG.info("Last sample: %s", self.last_sample)
            return
        if val < pval:
            # this is supposed to handle counter wrap around
            # see https://collectd.org/wiki/index.php/Data_source
            LOG.debug("COUNTER wrap-around for: %s:%s (%s -> %s)",
                      host, name, pval, val)
            if pval < 0x100000000:
                val += 0x100000000  # 2**32
            else:
                val += 0x10000000000000000  # 2**64
        return float(val - pval) / (time - ptime)

    def _calc_derive(self, host, name, val, time):
        """
        Calculating derived values
        """
        key = (host, name)
        if key not in self.prev_samples:
            self.prev_samples[key] = (val, time)
            return
        pval, ptime = self.prev_samples[key]
        self.prev_samples[key] = (val, time)
        if time <= ptime:
            LOG.debug("Invalid DERIVE update for: %s:%s", key[0], key[1])
            LOG.debug("Last sample: %s", self.last_sample)
            return
        return float(abs(val - pval)) / (time - ptime)

    def _calc_absolute(self, host, name, val, time):
        """
        Calculating absolute values
        """
        key = (host, name)
        if key not in self.prev_samples:
            self.prev_samples[key] = (val, time)
            return
        _, ptime = self.prev_samples[key]
        self.prev_samples[key] = (val, time)
        if time <= ptime:
            LOG.error("Invalid ABSOLUTE update for: %s:%s", key[0], key[1])
            LOG.info("Last sample: %s", self.last_sample)
            return
        return float(val) / (time - ptime)


class CollectDServer(UDPServer):
    """Single processes CollectDServer"""

    def __init__(self, queue):
        super(CollectDServer, self).__init__(settings.getValue('COLLECTD_IP'),
                                             settings.getValue('COLLECTD_PORT'))
        self.handler = CollectDHandler()
        self.queue = queue

    def handle(self, data, addr):
        for sample in self.handler.parse(data):
            self.queue.put(sample)
        return True

    def pre_shutdown(self):
        LOG.info("Sutting down CollectDServer")


def get_collectd_server(queue):
    """Get the collectd server """
    server = CollectDServer
    return server(queue)