commander.py

#!/usr/bin/env python

"""Cyber Hygiene commander.
Feeds scanners jobs, processes output, and stores results in database.

Usage:
  commander [options] <working-dir>
  commander (-h | --help)
  commander --version

Options:
  -b --background                Run in background (daemonize).
  -d --debug                     Enable debug logging.
  -l --stdout-log                Log to standard out.
  -s SECTION --section=SECTION   Configuration section to use.

"""


# easy-installs
# fabric
# python-daemon

from collections import defaultdict
import logging
import os
import Queue
import random
import shutil
import signal
import sys
import threading
import time
import traceback
from ConfigParser import SafeConfigParser

import daemon
from docopt import docopt
import lockfile
from fabric import operations
from fabric.api import task, run, env
from fabric.network import disconnect_all, normalize
from fabric.state import connections
from fabric.tasks import Task, execute

from cyhy.core import *
from cyhy.core.common import DEFAULT_OWNER, SCAN_TYPE, STAGE
from cyhy.db import CHDatabase, database
from cyhy.util import setup_logging

from job_sink import NmapSink, NessusSink, TryAgainSink, NoOpSink
from job_source import DirectoryJobSource, DatabaseJobSource

# fabric configuration
env.command_timeout = 60
env.keepalive = 30
env.use_ssh_config = True

# remote files
DONE_DIR = "runner/done"
DONE_FILE = ".done"
READY_FILE = ".ready"
RUNNING_DIR = "runner/running"

# local files
CONFIG_FILENAME = "/etc/cyhy/commander.conf"
DEFAULT_LOGGER_LEVEL = logging.INFO
DROP_DIR = "drop"
FAILED_DIR = "failed"
LOCK_FILENAME = "cyhy-commander"
LOG_FILE = "/var/log/cyhy/commander.log"
LOGGER_FORMAT = "%(asctime)-15s %(levelname)s %(name)s - %(message)s"
PUSHED_DIR = "pushed"
STOP_FILE = "stop"
SUCCESS_DIR = "done"

# local job files
me = os.path.realpath(__file__)
myDir = os.path.dirname(me)
jobsDir = os.path.join(myDir, "jobs")
BASESCAN_JOB_FILE = os.path.join(jobsDir, "basescan.sh")
NETSCAN1_JOB_FILE = os.path.join(jobsDir, "netscan1.sh")
NETSCAN2_JOB_FILE = os.path.join(jobsDir, "netscan2.sh")
PORTSCAN_JOB_FILE = os.path.join(jobsDir, "portscan.sh")
SLEEP_JOB_FILE = os.path.join(jobsDir, "rand-sleep.py")
VULNSCAN_JOB_FILE = os.path.join(jobsDir, "vulnscan.py")

# config file
DATABASE_NAME = "database-name"
DATABASE_URI = "database-uri"
DEFAULT = "DEFAULT"
DEFAULT_SCHEDULER = "default-scheduler"
DEFAULT_SECTION = "default-section"
JOB_PROCESSING_THREADS = "job-processing-threads"
JOBS_PER_NESSUS_HOST = "jobs-per-nessus-host"
JOBS_PER_NMAP_HOST = "jobs-per-nmap-host"
KEEP_FAILURES = "keep-failures"
KEEP_SUCCESSES = "keep-successes"
NESSUS_HOSTS = "nessus-hosts"
NEXT_SCAN_LIMIT = "next-scan-limit"
NMAP_HOSTS = "nmap-hosts"
POLL_INTERVAL = "poll-interval"
PRODUCTION_SECTION = "production"
SHUTDOWN_WHEN_IDLE = "shutdown-when-idle"
TEST_MODE = "test-mode"
TESTING_PURGE_SECTION = "testing-purge"
TESTING_SECTION = "testing"

# TODO eventual config options
IPS_PER_BASESCAN_JOB = 32
IPS_PER_NETSCAN1_JOB = 128
IPS_PER_NETSCAN2_JOB = 128 / 2
IPS_PER_PORTSCAN_JOB = 32 / 4
IPS_PER_VULNSCAN_JOB = 4
RANDOMIZE_SOURCES = True

NESSUS_WORKGROUP = "nessus"
NMAP_WORKGROUP = "nmap"

# The number of exceptions to allow before putting a host on "cooldown".
#
# A host on cooldown is removed from any work group lists it is on for the
# duration of its cooldown period. This means that the existing connection is
# closed, and no attempts are made to interact with the host until the below
# cooldown duration has expired. Once it has, the host is restored to the
# list(s) it was removed from and normal operations for that host continue.
NUM_EXCEPTIONS_ALLOWED = 2
# How long should a host remain on cooldown
#
# This value is in seconds, but it is set up to use minutes for granularity. The
# second value should be changed to modify the cooldown duration.
COOLDOWN_DURATION = 60 * 30


class Commander(object):
    def __init__(self, config_section=None, debug_logging=False, console_logging=False):
        # Set up logging first in order to log any errors as soon as possible.
        self.__logger = logging.getLogger(__name__)
        self.__setup_logging(debug_logging, console_logging)

        self.__all_hosts_idle = False
        self.__config_section = config_section
        self.__db = None
        self.__failed_job_queue = None
        self.__failure_sinks = []
        self.__host_exceptions = defaultdict(lambda: 0)
        self.__hosts_on_cooldown = []
        self.__is_processing_jobs = True
        self.__is_running = True
        self.__job_processing_sleep_duration = 1
        self.__keep_failures = False
        self.__keep_successes = False
        self.__log_output_sleep_duration = 10
        self.__nessus_sources = []
        self.__next_scan_limit = 2000
        self.__nmap_sources = []
        self.__queue_monitor_output_lock = threading.Lock()
        self.__setup_directories()
        self.__shutdown_when_idle = False
        self.__success_sinks = []
        self.__successful_job_queue = None
        self.__test_mode = False

    def __setup_logging(self, debug_logging, console_logging):
        # get default logging setup
        if debug_logging:
            level = logging.DEBUG
        else:
            level = DEFAULT_LOGGER_LEVEL

        if console_logging:
            setup_logging(level, console=True)
        else:
            setup_logging(level, filename=LOG_FILE)

        # This is only output if debug in enabled.  Skipped outwise.
        self.__logger.debug("Debug logging enabled")

    def __setup_directories(self):
        for directory in (SUCCESS_DIR, PUSHED_DIR, FAILED_DIR):
            if not os.path.exists(directory):
                self.__logger.info('Creating directory "%s".' % (directory))
                os.makedirs(directory)

    def __setup_db(self, db_name, uri):
        self.__db = database.db_from_connection(uri, db_name)
        self.__ch_db = CHDatabase(self.__db, next_scan_limit=self.__next_scan_limit)

    def __setup_sources(self):
        if self.__test_mode:
            self.__nmap_sources.append(
                DatabaseJobSource(
                    SLEEP_JOB_FILE,
                    self.__db,
                    job_type=STAGE.NETSCAN1,
                    count=IPS_PER_NETSCAN1_JOB,
                )
            )
            self.__nmap_sources.append(
                DatabaseJobSource(
                    SLEEP_JOB_FILE,
                    self.__db,
                    job_type=STAGE.NETSCAN2,
                    count=IPS_PER_NETSCAN2_JOB,
                )
            )
            self.__nmap_sources.append(
                DatabaseJobSource(
                    SLEEP_JOB_FILE,
                    self.__db,
                    job_type=STAGE.PORTSCAN,
                    count=IPS_PER_PORTSCAN_JOB,
                )
            )
            self.__nessus_sources.append(
                DatabaseJobSource(
                    SLEEP_JOB_FILE,
                    self.__db,
                    job_type=STAGE.VULNSCAN,
                    count=IPS_PER_VULNSCAN_JOB,
                )
            )
            # self.__nmap_sources.append(DatabaseJobSource(SLEEP_JOB_FILE, self.__db, job_type=STAGE.BASESCAN, count=IPS_PER_BASESCAN_JOB))
        else:
            self.__nessus_sources.append(DirectoryJobSource(DROP_DIR))
            self.__nmap_sources.append(
                DatabaseJobSource(
                    NETSCAN1_JOB_FILE,
                    self.__db,
                    job_type=STAGE.NETSCAN1,
                    count=IPS_PER_NETSCAN1_JOB,
                )
            )
            self.__nmap_sources.append(
                DatabaseJobSource(
                    NETSCAN2_JOB_FILE,
                    self.__db,
                    job_type=STAGE.NETSCAN2,
                    count=IPS_PER_NETSCAN2_JOB,
                )
            )
            self.__nmap_sources.append(
                DatabaseJobSource(
                    PORTSCAN_JOB_FILE,
                    self.__db,
                    job_type=STAGE.PORTSCAN,
                    count=IPS_PER_PORTSCAN_JOB,
                )
            )
            self.__nessus_sources.append(
                DatabaseJobSource(
                    VULNSCAN_JOB_FILE,
                    self.__db,
                    job_type=STAGE.VULNSCAN,
                    count=IPS_PER_VULNSCAN_JOB,
                )
            )
            # self.__nmap_sources.append(DatabaseJobSource(BASESCAN_JOB_FILE, self.__db, job_type=STAGE.BASESCAN, count=IPS_PER_BASESCAN_JOB))

    def __setup_sinks(self):
        if self.__test_mode:
            noop_sink = NoOpSink(self.__db)
            self.__success_sinks.append(noop_sink)
        else:
            netscan1_sink = NmapSink(self.__db, STAGE.NETSCAN1)
            netscan2_sink = NmapSink(self.__db, STAGE.NETSCAN2)
            portscan_sink = NmapSink(self.__db, STAGE.PORTSCAN)
            vulnscan_sink = NessusSink(self.__db)
            # baseline_sink = NmapSink(self.__db, STAGE.BASESCAN)
            self.__success_sinks.extend(
                (netscan1_sink, netscan2_sink, portscan_sink, vulnscan_sink)
            )
        self.__failure_sinks = [TryAgainSink(self.__db)]

    @task
    def __done_jobs(self):
        try:
            output = run("ls %s" % DONE_DIR)
            if output.failed:
                self.__logger.warning(
                    'Unable to get listing of "%s" on %s' % (DONE_DIR, env.host_string)
                )
                output = ""
            doneJobs = output.split()
            for job in doneJobs:
                jobPath = os.path.join(DONE_DIR, job)
                output = run("ls -a %s" % (jobPath))
                jobContents = output.split()
                if DONE_FILE in jobContents:
                    self.__logger.info(
                        "%s is ready for pickup on %s" % (job, env.host_string)
                    )
                    donePath = os.path.join(jobPath, DONE_FILE)
                    exitCode = run("cat %s" % donePath)
                    if exitCode == "0":
                        destDir = SUCCESS_DIR
                    else:
                        destDir = FAILED_DIR
                        self.__logger.warning(
                            "%s had a non-zero exit code: %s" % (job, exitCode)
                        )

                    paths = operations.get(jobPath, destDir)
                    if len(paths.failed) == 0:
                        self.__logger.info(
                            "%s was copied successfully from %s to %s"
                            % (job, env.host_string, destDir)
                        )
                        # remove remote dir
                        run("rm -rf %s" % jobPath)
                        self.__logger.info(
                            "%s was removed from %s" % (job, env.host_string)
                        )
                    local_job_path = os.path.join(destDir, job)

                    if destDir == SUCCESS_DIR:
                        self.__successful_job_queue.put(local_job_path)
                    else:
                        self.__failed_job_queue.put(local_job_path)

                else:
                    self.__logger.warning(
                        "%s is not ready for pickup on %s" % (job, env.host_string)
                    )
        except Exception as e:
            self.__logger.error(
                "Exception when retrieving done jobs from %s" % env.host_string
            )
            self.__logger.error(e)
            self.__host_exceptions[env.host_string] += 1

    @task
    def __running_job_count(self):
        try:
            output = run("ls %s" % RUNNING_DIR)
            if output.failed:
                self.__logger.warning(
                    'Unable to get listing of "%s" on %s'
                    % (RUNNING_DIR, env.host_string)
                )
                return None
            runningJobs = output.split()
            count = len(runningJobs)
            return count
        except Exception as e:
            self.__logger.error(
                "Exception when retrieving running job count from %s" % env.host_string
            )
            self.__logger.error(e)
            self.__host_exceptions[env.host_string] += 1

    @task
    def __push_job(self, job_path):
        try:
            paths = operations.put(job_path, RUNNING_DIR)
            if len(paths.failed) == 0:
                self.__logger.info(
                    "%s was pushed successfully to %s" % (job_path, env.host_string)
                )
                job_name = os.path.basename(job_path)
                run("touch %s" % os.path.join(RUNNING_DIR, job_name, READY_FILE))
                self.__move_to_pushed(job_path)
            else:
                self.__logger.error(
                    "Error pushing %s to host %s" % (job_path, env.host_string)
                )
        except Exception as e:
            self.__logger.error(
                "Exception when pushing %s to host %s" % (job_path, env.host_string)
            )
            self.__logger.error(e)
            self.__host_exceptions[env.host_string] += 1

    def __unique_filename(self, path):
        if not os.path.exists(path):
            return path
        new_name = "%s.%d" % (os.path.basename(path), int(time.time() * 1000000))
        new_path = os.path.join(os.path.dirname(path), new_name)
        return new_path

    def __move_to_pushed(self, job_path):
        if not self.__test_mode:
            shutil.rmtree(job_path)
            self.__logger.info("%s deleted" % job_path)
        else:
            dest = os.path.join(PUSHED_DIR, os.path.basename(job_path))
            dest = self.__unique_filename(dest)
            shutil.move(job_path, dest)
            self.__logger.info("%s moved locally to %s" % (job_path, dest))

    def __lowest_host(self, counts):
        lowest_count = None
        lowest_host = None
        for (host, count) in counts.items():
            if count != None and (lowest_count == None or count < lowest_count):
                lowest_host = host
                lowest_count = count
        return lowest_host

    def __all_idle(self, counts):
        for (host, count) in counts.items():
            if count > 0:
                self.__all_hosts_idle = False
                return False
        if not self.__all_hosts_idle:
            self.__all_hosts_idle = True
            self.__logger.info("All hosts are now idle.")
            if self.__shutdown_when_idle:
                self.__logger.warning("Shutting down since all hosts are idle.")
                self.__is_running = False

    def __job_from_sources(self, sources):
        job = None
        if RANDOMIZE_SOURCES:
            random.shuffle(sources)
        for source in sources:
            self.__logger.debug("Checking %s for a job." % source)
            job = source.get_job()
            if job != None:
                self.__logger.info("Acquired a job from %s" % source)
                break
            self.__logger.debug("No available jobs returned by %s" % source)
        return job

    def __fill_hosts(self, counts, sources, workgroup_name, jobs_per_host):
        while True:
            lowest_host = self.__lowest_host(counts)
            if counts[lowest_host] >= jobs_per_host:
                self.__logger.debug("All %s hosts are full" % workgroup_name)
                break  # everyone is full
            job_path = self.__job_from_sources(sources)
            if job_path == None:
                self.__logger.debug(
                    "Not enough work available to fill %s hosts" % workgroup_name
                )
                break  # no more work to do
            execute(self.__push_job, self, job_path, hosts=[lowest_host])
            counts[lowest_host] += 1

    def __monitor_job_queues(self):
        # Output the number of jobs that are not done for each queue every
        # self.__log_output_sleep_duration seconds while work is on the queues.
        while self.__is_processing_jobs:
            with self.__queue_monitor_output_lock:
                self.__logger.debug(
                    "%d unfinished jobs in the successful job queue"
                    % self.__successful_job_queue.unfinished_tasks
                )
                self.__logger.debug(
                    "%d unfinished jobs in the failed job queue"
                    % self.__failed_job_queue.unfinished_tasks
                )
            time.sleep(self.__log_output_sleep_duration)

    def __process_queued_jobs(self):

        # define an inner function to process jobs
        def process_job_from_queue(target_job_queue, job_processing_function):
            """Helper function to process jobs from a queue.

            Args:
                target_job_queue (Queue.Queue): The queue to get a job to process.
                job_processing_function (callable): The function used to process a job.

            Returns:
                The job path that was processed or None if the queue was empty.
            """
            job_path = None

            # check the successful jobs queue
            try:
                job_path = target_job_queue.get(timeout=1)
            except Queue.Empty:
                return job_path

            try:
                job_processing_function(job_path)
            except Exception, e:
                self.__logger.critical(e)
                self.__logger.critical(traceback.format_exc())

            # report task completion no matter what so the queue can be joined
            target_job_queue.task_done()

            # return path of the job that was processed
            return job_path

        # run as long as the commander is processing jobs
        while self.__is_processing_jobs:
            # process successful job
            job_processing_results = process_job_from_queue(
                self.__successful_job_queue, self.__process_successful_job
            )

            # process failed job if a successful job was not processed
            if job_processing_results is None:
                job_processing_results = process_job_from_queue(
                    self.__failed_job_queue, self.__process_failed_job
                )

            # sleep if both queues are empty
            if job_processing_results is None:
                time.sleep(self.__job_processing_sleep_duration)

    def __process_successful_job(self, job_path):
        # Get the name of the current thread
        thread_name = threading.current_thread().name

        for sink in self.__success_sinks:
            if sink.can_handle(job_path):
                self.__logger.info(
                    "[%s] Processing %s with %s" % (thread_name, job_path, sink)
                )
                sink.handle(job_path)
                self.__logger.info("[%s] Processing completed" % thread_name)
                if not self.__test_mode and not self.__keep_successes:
                    shutil.rmtree(job_path)
                    self.__logger.info("[%s] %s deleted" % (thread_name, job_path))
                return
        self.__logger.warning(
            "[%s] No handler was able to process %s" % (thread_name, job_path)
        )

    def __process_failed_job(self, job_path):
        # Get the name of the current thread
        thread_name = threading.current_thread().name

        for sink in self.__failure_sinks:
            if sink.can_handle(job_path):
                self.__logger.warning(
                    "[%s] Processing %s with %s" % (thread_name, job_path, sink)
                )
                sink.handle(job_path)
                self.__logger.info("[%s] Processing completed" % thread_name)
                if not self.__test_mode and not self.__keep_failures:
                    shutil.rmtree(job_path)
                    self.__logger.info("[%s] %s deleted" % (thread_name, job_path))
                return
        self.__logger.warning(
            "[%s] No handler was able to process %s" % (thread_name, job_path)
        )

    def handle_term(self, signal, frame):
        self.__logger.warning(
            "Received signal %d.  Shutting down after this work cycle completes."
            % signal
        )
        self.__is_running = False

    def __check_stop_file(self):
        if os.path.exists(STOP_FILE):
            self.__logger.warning(
                "Stop file found.  Shutting down after this work cycle completes."
            )
            os.remove(STOP_FILE)
            self.__is_running = False

    def __check_database_pause(self):
        while self.__ch_db.should_commander_pause() and self.__is_running:
            self.__logger.info("Commander is paused due to database request.")
            time.sleep(self.__log_output_sleep_duration)
            self.__check_stop_file()

    def __write_config(self):
        config = SafeConfigParser()
        config.set(None, DATABASE_URI, "mongodb://localhost:27017/")
        config.set(None, JOBS_PER_NMAP_HOST, "8")
        config.set(None, JOBS_PER_NESSUS_HOST, "8")
        config.set(None, JOB_PROCESSING_THREADS, "4")
        config.set(None, POLL_INTERVAL, "30")
        config.set(None, NEXT_SCAN_LIMIT, "2000")
        config.set(None, DEFAULT_SECTION, TESTING_SECTION)
        config.set(None, TEST_MODE, "false")
        config.set(None, KEEP_FAILURES, "false")
        config.set(None, KEEP_SUCCESSES, "false")
        config.set(None, SHUTDOWN_WHEN_IDLE, "false")
        config.set(None, DEFAULT_SCHEDULER, "PERSISTENT1")
        config.add_section(TESTING_SECTION)
        config.set(TESTING_SECTION, NMAP_HOSTS, "comma,separated,list")
        config.set(TESTING_SECTION, NESSUS_HOSTS, "comma,separated,list")
        config.set(TESTING_SECTION, DATABASE_NAME, "test_database")
        config.set(TESTING_SECTION, TEST_MODE, "true")
        config.add_section(TESTING_PURGE_SECTION)
        config.set(TESTING_PURGE_SECTION, JOBS_PER_NMAP_HOST, "0")
        config.set(TESTING_PURGE_SECTION, JOBS_PER_NESSUS_HOST, "0")
        config.set(TESTING_PURGE_SECTION, SHUTDOWN_WHEN_IDLE, "true")
        config.set(TESTING_PURGE_SECTION, NMAP_HOSTS, "comma,separated,list")
        config.set(TESTING_PURGE_SECTION, NESSUS_HOSTS, "comma,separated,list")
        config.set(TESTING_PURGE_SECTION, DATABASE_NAME, "test_database")
        config.set(TESTING_PURGE_SECTION, TEST_MODE, "true")
        config.add_section(PRODUCTION_SECTION)
        config.set(PRODUCTION_SECTION, NMAP_HOSTS, "comma,separated,list")
        config.set(PRODUCTION_SECTION, NESSUS_HOSTS, "comma,separated,list")
        config.set(PRODUCTION_SECTION, DATABASE_NAME, "test_database")
        with open(CONFIG_FILENAME, "wb") as config_file:
            config.write(config_file)

    def __read_config(self):
        config = SafeConfigParser()
        config.read([CONFIG_FILENAME])
        return config

    def __setup_default_owner(self, scheduler):
        """Ensures that a RequestDoc exists in the database for the default owner.

        This function checks if a RequestDoc for the default owner exists, and
        if not, creates one with default values.  This function also enables
        scanning for the default owner and sets the scheduler as specified.

        The default owner owns all "ownerless" HostDocs.  The default owner's
        RequestDoc does not have a valid list of networks (IP addresses), but it
        does have scan windows and concurrency settings.  Those "ownerless"
        HostDocs are created when a CyHy entity has a hostname that resolves to
        IP addresses that are not already owned by a CyHy entity.  This is how
        we account for cases where an entity owns a hostname, but not
        necessarily the IP addresses that it resolves to.

        Args:
            scheduler (str): The scheduler value to assign to the default
            owner's RequestDoc.

        Returns:
            None
        """
        if not self.__db.RequestDoc.get_by_owner(DEFAULT_OWNER):
            self.__logger.info(
                "%s request document does not exist; creating..." % DEFAULT_OWNER
            )
            # Create a new request document populated with default values
            request = self.__db.RequestDoc()
            # Customize request document for the default owner
            request["_id"] = DEFAULT_OWNER
            request["agency"]["acronym"] = DEFAULT_OWNER
            request["agency"]["name"] = "Default CyHy system owner"
            # Remove the location field; it is not required here
            request["agency"].pop("location")
            # Enable scanning for default owner
            request["scan_types"] = [SCAN_TYPE.CYHY]
            request["scheduler"] = scheduler
            request.save()
            self.__logger.info("%s request document created" % DEFAULT_OWNER)

    def do_work(self):
        env.warn_only = True
        self.__logger.info("Starting up.")
        self.__setup_directories()

        # process configuration
        if not os.path.exists(CONFIG_FILENAME):
            print >>sys.stderr, 'Configuration file not found: "%s"' % CONFIG_FILENAME
            self.__write_config()
            print >>sys.stderr, "A default configuration file was created in the working directory."
            print >>sys.stderr, "Please edit and relaunch."
            self.__logger.error("Configuration file not found. Exiting.")
            sys.exit(-1)

        config = self.__read_config()
        if self.__config_section == None:
            config_section = config.get(DEFAULT, DEFAULT_SECTION)
        else:
            config_section = self.__config_section
        self.__logger.info('Reading configuration section: "%s"' % config_section)

        nmap_hosts = config.get(config_section, NMAP_HOSTS).split(",")
        nessus_hosts = config.get(config_section, NESSUS_HOSTS).split(",")
        # remove any duplicates
        nmap_hosts = list(set(nmap_hosts))
        nessus_hosts = list(set(nessus_hosts))
        # clean up the host lists and sort them
        nmap_hosts = sorted([h.strip() for h in nmap_hosts if h])
        nessus_hosts = sorted([h.strip() for h in nessus_hosts if h])
        # clean up empty lists from config
        if not nmap_hosts:
            nmap_hosts = None
        if not nessus_hosts:
            nessus_hosts = None

        self.__logger.info("nmap hosts: %s" % nmap_hosts)
        self.__logger.info("nessus hosts: %s" % nessus_hosts)
        jobs_per_nmap_host = config.getint(config_section, JOBS_PER_NMAP_HOST)
        self.__logger.info("Jobs per nmap host: %d", jobs_per_nmap_host)
        jobs_per_nessus_host = config.getint(config_section, JOBS_PER_NESSUS_HOST)
        self.__logger.info("Jobs per nessus host: %d", jobs_per_nessus_host)
        self.__next_scan_limit = config.getint(config_section, NEXT_SCAN_LIMIT)
        self.__logger.info("Next scan fetch limit: %d", self.__next_scan_limit)
        self.__poll_interval = config.getint(config_section, POLL_INTERVAL)
        self.__logger.info("Poll interval: %d", self.__poll_interval)
        db_name = config.get(config_section, DATABASE_NAME)
        db_uri = config.get(config_section, DATABASE_URI)
        self.__setup_db(db_name, db_uri)
        self.__logger.info("Database: %s", self.__db)
        self.__test_mode = config.getboolean(config_section, TEST_MODE)
        self.__logger.info("Test mode: %s", self.__test_mode)
        self.__keep_failures = config.getboolean(config_section, KEEP_FAILURES)
        job_processing_thread_count = config.getint(
            config_section, JOB_PROCESSING_THREADS
        )
        self.__logger.info(
            "Number of job processing threads: %d", job_processing_thread_count
        )
        self.__logger.info("Keep failed jobs: %s", self.__keep_failures)
        self.__keep_successes = config.getboolean(config_section, KEEP_SUCCESSES)
        self.__logger.info("Keep successful jobs: %s", self.__keep_successes)
        self.__shutdown_when_idle = config.getboolean(
            config_section, SHUTDOWN_WHEN_IDLE
        )
        self.__logger.info("Idle shutdown: %s", self.__shutdown_when_idle)
        self.__logger.info('Default owner: "%s"' % DEFAULT_OWNER)
        default_scheduler = config.get(config_section, DEFAULT_SCHEDULER)
        self.__logger.info('Default scheduler: "%s"' % default_scheduler)
        self.__setup_default_owner(default_scheduler)
        self.__setup_sources()
        self.__setup_sinks()

        self.__successful_job_queue = Queue.Queue()
        self.__failed_job_queue = Queue.Queue()

        # spin up the thread pool to process retrieved work
        job_processing_threads = []
        for t in range(job_processing_thread_count):
            job_processing_thread = threading.Thread(
                name="JobProcessor-%d" % t, target=self.__process_queued_jobs
            )
            job_processing_threads.append(job_processing_thread)
            try:
                job_processing_thread.start()
            except Exception as e:
                self.__logger.error("Unable to start job processing thread #%s", t)
                self.__logger.error(e)
                # bail out
                self.__logger.critical(
                    "Shutting down due to inability to start job processing threads."
                )
                self.__is_running = False

        # spin up a thread to output queue load information
        self.__queue_monitor_output_lock.acquire()
        job_queue_monitor_thread = threading.Thread(
            name="QueueMonitor", target=self.__monitor_job_queues
        )
        try:
            job_queue_monitor_thread.start()
        except Exception as e:
            self.__logger.error("Unable to start job queue monitoring thread")
            self.__logger.error(e)
            # bail out
            self.__logger.critical(
                "Shutting down due to inability to start queue monitoring thread."
            )
            self.__is_running = False

        # pairs of hosts and job sources
        work_groups = (
            (NMAP_WORKGROUP, nmap_hosts, self.__nmap_sources, jobs_per_nmap_host),
            (
                NESSUS_WORKGROUP,
                nessus_hosts,
                self.__nessus_sources,
                jobs_per_nessus_host,
            ),
        )

        # main work loop
        while self.__is_running:
            try:
                # record time at start of duty cycle
                cycle_start_time = time.time()
                next_cycle_start_time = cycle_start_time + self.__poll_interval

                # check for hosts that are coming off of cooldown
                self.__logger.debug("Checking for hosts to bring off of cooldown")
                # we don't want to modify the list while we are iterating, so we
                # iterate through a copy and remove from the original
                for host_info in self.__hosts_on_cooldown[:]:
                    cooldown_end = host_info["cooldown_start"] + COOLDOWN_DURATION
                    if time.time() >= cooldown_end:
                        try:
                            # Manually set the appropriate environment value
                            env.host_string = host_info["host"]
                            # Manually re-connect to the host
                            for cache_key in host_info["cache_keys"]:
                                connections.connect(cache_key)
                        except Exception as e:
                            self.__logger.error(
                                "Unable to reconnect to '%s'" % host_info["host"]
                            )
                            self.__logger.error(e)
                            continue
                        for group in host_info["work_groups"]:
                            work_groups[group][1].append(host_info["host"])
                            work_groups[group][1].sort()
                        self.__hosts_on_cooldown.remove(host_info)
                        self.__logger.debug(
                            "Host '%s' has been put back into rotation"
                            % host_info["host"]
                        )
                    else:
                        self.__logger.debug(
                            "Host '%s' is out of rotation until %s"
                            % (
                                host_info["host"],
                                # output an ISO 8601 human readable time
                                time.strftime(
                                    "%Y-%m-%dT%H:%M:%S", time.localtime(cooldown_end)
                                ),
                            )
                        )

                # check for hosts that have had multiple exceptions
                self.__logger.debug("Checking for hosts with too many exceptions")
                for (host, count) in self.__host_exceptions.items():
                    if count > NUM_EXCEPTIONS_ALLOWED:
                        groups = []
                        for (index, group) in enumerate(work_groups):
                            if host in group[1]:
                                groups.append(index)
                                group[1].remove(host)

                        info_dict = {
                            "host": host,
                            "cooldown_start": time.time(),
                            "work_groups": groups,
                        }
                        self.__host_exceptions[host] = 0
                        cache_keys = []
                        for cache_key in connections.keys():
                            cuser, chost, cport = normalize(cache_key)
                            if chost == host:
                                connections[cache_key].close()
                                cache_keys.append(cache_key)
                        info_dict["cache_keys"] = cache_keys
                        self.__hosts_on_cooldown.append(info_dict)
                        self.__logger.debug(
                            "Host '%s' has been removed from rotation" % host
                        )

                # process anything that has completed
                self.__logger.debug(
                    "Checking remotes for completed jobs to download and process"
                )
                self.__queue_monitor_output_lock.release()
                for (workgroup_name, hosts, sources, jobs_per_host) in work_groups:
                    if hosts == None:
                        continue
                    execute(self.__done_jobs, self, hosts=hosts)

                # wait for work to process
                self.__logger.debug("Waiting for completed jobs to be processed.")
                self.__successful_job_queue.join()
                self.__failed_job_queue.join()
                self.__queue_monitor_output_lock.acquire()

                # check for scheduled hosts
                self.__logger.debug(
                    "Checking for scheduled DONE hosts to mark WAITING."
                )
                self.__ch_db.check_host_next_scans()

                # balance the number of hosts that are ready and running
                self.__logger.debug("Balancing READY status of hosts.")
                self.__ch_db.balance_ready_hosts()

                # push out new work and count
                self.__logger.debug("Checking sources for new jobs")
                all_workgroup_counts = {}  # track counts from each work_group
                for (workgroup_name, hosts, sources, jobs_per_host) in work_groups:
                    if hosts == None:
                        continue
                    counts = execute(self.__running_job_count, self, hosts=hosts)
                    self.__fill_hosts(counts, sources, workgroup_name, jobs_per_host)
                    all_workgroup_counts.update(counts)

                # check to see if all host are idle and log it
                self.__all_idle(all_workgroup_counts)
                self.__check_stop_file()
                if self.__is_running:
                    now = time.time()
                    if now < next_cycle_start_time:
                        sleep_time = next_cycle_start_time - now
                        self.__logger.debug(
                            "Sleeping for %1.1f seconds.\n\n\n" % sleep_time
                        )
                        time.sleep(sleep_time)
                    else:
                        self.__logger.debug(
                            "No time to sleep.  Last cycle took %1.1f seconds.\n\n\n"
                            % (now - cycle_start_time)
                        )
                self.__check_stop_file()
                self.__check_database_pause()
            except Exception, e:
                self.__logger.critical(e)
                self.__logger.critical(traceback.format_exc())

        # signal job processing threads to exit once they have finished all
        # queued work
        self.__is_processing_jobs = False
        self.__queue_monitor_output_lock.release()

        # wait for the job processing threads to exit
        for job_processing_thread in job_processing_threads:
            job_processing_thread.join()

        # wait for the job queue monitoring thread to exit
        job_queue_monitor_thread.join()

        self.__logger.info("Shutting down.")
        disconnect_all()


def main():
    args = docopt(__doc__, version="v1.1.0")
    workingDir = os.path.join(os.getcwd(), args["<working-dir>"])
    if not os.path.exists(workingDir):
        print >>sys.stderr, 'Working directory "%s" does not exist.  Attempting to create...' % workingDir
        os.mkdir(workingDir)
    os.chdir(workingDir)
    lock = lockfile.LockFile(os.path.join(workingDir, LOCK_FILENAME), timeout=0)
    if lock.is_locked():
        print >>sys.stderr, "Cannot start.  There is already a cyhy-commander executing in this working directory."
        sys.exit(-1)

    commander = Commander(args["--section"], args["--debug"], args["--stdout-log"])

    if args["--background"]:
        context = daemon.DaemonContext(
            working_directory=workingDir, umask=0002, pidfile=lock
        )
        context.signal_map = {
            signal.SIGTERM: commander.handle_term,
            signal.SIGCHLD: signal.SIG_IGN,
        }
        with context:
            commander.do_work()
    else:
        signal.signal(signal.SIGTERM, commander.handle_term)
        signal.signal(signal.SIGINT, commander.handle_term)
        commander.do_work()


if __name__ == "__main__":
    main()