Observium_CE/poller-wrapper.py

1178 lines
43 KiB
Python
Executable File

#!/usr/bin/env python
"""
Observium
This file is part of Observium.
@package observium
@subpackage poller
@copyright (C) 2013-2014 Job Snijders, (C) 2014-2023 Observium Limited
"""
"""
observium-wrapper A small tool which wraps around the Observium poller/discovery/billing
and tries to guide the relevant process with a more modern
approach with a Queue and workers
Original: Job Snijders <job.snijders@atrato.com>
Rewritten: Most code parts rewritten since 2014 by Observium developers
Usage: This program accepts process name as first command line argument,
that should run simultaneously. For other options see help -h
In /etc/cron.d/observium replace this (or the equivalent) poller entry:
*/5 * * * * root /opt/observium/poller.php -h all >> /dev/null 2>&1
with something like this:
*/5 * * * * root /opt/observium/observium-wrapper poller >> /dev/null 2>&1
In /etc/cron.d/observium replace this (or the equivalent) discovery entries:
34 */6 * * * root /opt/observium/discovery.php -h all >> /dev/null 2>&1
*/5 * * * * root /opt/observium/discovery.php -h new >> /dev/null 2>&1
with something like this:
34 */6 * * * root /opt/observium/observium-wrapper discovery >> /dev/null 2>&1
*/5 * * * * root /opt/observium/observium-wrapper discovery --host new >> /dev/null 2>&1
In /etc/cron.d/observium replace this (or the equivalent) billing entries:
*/5 * * * * root /opt/observium/poll-billing.php >> /dev/null 2>&1
with something like this:
*/5 * * * * root /opt/observium/observium-wrapper billing >> /dev/null 2>&1
Ubuntu Linux: apt-get install python-mysqldb
New Ubuntu: sudo apt install python3-pymysql
RHEL/CentOS: yum install MySQL-python (Requires the EPEL repo!)
FreeBSD: cd /usr/ports/*/py-MySQLdb && make install clean
Tested on: Python 2.7.3 / PHP 5.3.10-1ubuntu3.4 / Ubuntu 12.04 LTS
Python 3.2.3 / PHP 5.3.27-1ubuntu3.6 / Ubuntu 12.04 LTS
Python 2.7.6 / PHP 5.5.9-1ubuntu4.14 / Ubuntu 14.04 LTS
Python 3.4.3 / PHP 5.5.9-1ubuntu4.14 / Ubuntu 14.04 LTS
Python 2.7.12 / PHP 7.0.13 / Ubuntu 16.04 LTS
Python 3.5.2 / PHP 7.0.13 / Ubuntu 16.04 LTS
Python 3.6.6 / PHP 7.2.10 / Ubuntu 18.04 LTS
"""
"""
Import required modules
"""
try:
# Import time module first for more accurate start time
import time
# start time
s_time = time.time()
except ImportError:
print("ERROR: missing python module: time")
exit(2)
try:
# Required modules
import threading
import sys
import subprocess
import os
import json
import stat
except ImportError:
print("ERROR: missing one or more of the following python modules:")
print("threading, sys, subprocess, os, json, stat")
sys.exit(2)
def new_except_hook(exctype, value, traceback):
"""
Register global exepthook for ability stop execute wrapper by Ctrl+C
See: https://stackoverflow.com/questions/6598053/python-global-exception-handling
"""
if exctype == KeyboardInterrupt:
print("\n\nExiting by CTRL+C.\n\n")
sys.exit(2)
else:
sys.__excepthook__(exctype, value, traceback)
sys.excepthook = new_except_hook
class Colors:
HEADER = '\033[95m'
OKBLUE = '\033[94m'
OKGREEN = '\033[92m'
WARNING = '\033[93m'
FAIL = '\033[91m'
ENDC = '\033[0m'
BOLD = '\033[1m'
UNDERLINE = '\033[4m'
def logfile(log):
"""
Definition for writing msg to log file
"""
log = "[%s] %s(%s): %s/%s: %s\n" % (
time.strftime("%Y/%m/%d %H:%M:%S %z"), scriptname, os.getpid(), config['install_dir'], scriptname, log)
# https://jira.observium.org/browse/OBS-2631
# if the log file is a "character special device file" or a "FIFO (named pipe)" we must use mode 'w'
try:
fstat = os.stat(log_path).st_mode
if stat.S_ISCHR(fstat) or stat.S_ISFIFO(fstat):
fmode = 'w'
else:
fmode = 'a'
except OSError:
print("\nLog file %s doesn't exist.\n" % log_path)
return
try:
with open(log_path, fmode) as f:
f.write(log)
except IOError:
print("\nLog file %s is not writeable.\n" % log_path)
# base script name
scriptname = os.path.basename(sys.argv[0])
# major/minor python version: (2,7), (3,2), etc
python_version = sys.version_info[:2]
if python_version < (3, 0):
try:
import Queue
except ImportError:
print("ERROR: missing python module: Queue")
sys.exit(2)
try:
import MySQLdb
db_version = "MySQLdb " + MySQLdb.__version__
except ImportError:
print("ERROR: missing python module: MySQLdb")
print("On Ubuntu: apt-get install python-mysqldb")
print("On RHEL/CentOS: yum install MySQL-python")
print("On FreeBSD: cd /usr/ports/*/py-MySQLdb && make install clean")
sys.exit(2)
else:
try:
import queue as Queue
except ImportError:
print("ERROR: missing python module: queue")
sys.exit(2)
# MySQLdb not available for python3
# http://stackoverflow.com/questions/4960048/python-3-and-mysql
try:
import pymysql as MySQLdb
MySQLdb.install_as_MySQLdb()
import MySQLdb
db_version = "pymysql " + MySQLdb.__version__
except ImportError:
print("ERROR: missing python module: pymysql")
print(" On Ubuntu >= 16.04 and Debian >= 9.0:")
print(" sudo apt install python3-pymysql")
print(" On older Ubuntu/Debian:")
print(" sudo apt-get install python3-setuptools")
print(" sudo easy_install3 pip")
print(" sudo pip3 install PyMySQL")
# FIXME. I not know how install on RHEL and FreeBSD
print(" On other OSes install pip for python3 and then run from root user:")
print(" pip3 install PyMySQL")
# print(" On RHEL/CentOS: yum install MySQL-python")
# print(" On FreeBSD: cd /usr/ports/*/py-MySQLdb && make install clean")
sys.exit(2)
"""
Parse Arguments
Attempt to use argparse module. Probably want to use this moving forward
especially as more features want to be added to this wrapper.
and
Take the amount of threads we want to run in parallel from the commandline
if none are given or the argument was garbage, fall back to default of 8
"""
workers_arg = True # need for check compatibility with old syntax
try:
import argparse
parser = argparse.ArgumentParser(description='Process Wrapper for Observium')
parser.add_argument('process', nargs='?', default='poller',
help='Process name, one of \'poller\', \'discovery\', \'billing\'.')
# parser.add_argument('process', nargs='?', default='poller', choices=['poller', 'discovery', 'billing'], help='Process name, one of \'poller\', \'discovery\', \'billing\'.')
parser.add_argument('-w', '--workers', nargs='?', type=int, default=0,
help='Number of threads to spawn. Default: CPUs x 2')
parser.add_argument('--timeout', nargs='?', type=int, default=0,
help='Hard timeout for run poller/discovery for each device in seconds (greater than 90s).')
parser.add_argument('-p', '--poller_id', nargs='?', type=int, default=-1,
help='Specify poller_id for a partitioned poller. Usually not needed for partitioned operation.')
parser.add_argument('-s', '--stats', action='store_true', help='Store total polling times to RRD.', default=False)
parser.add_argument('-i', '--instances', nargs='?', type=int, default=-1, help='Process instances count.')
parser.add_argument('-n', '--number', nargs='?', type=int, default=-1,
help='Instance id (number), must start from 0 and to be less than instances count.')
parser.add_argument('-g', '--include-groups', nargs='*', type=int,
help='List of device group IDs to restrict polling/discovery these groups only. List is separated '
'by spaces (-g 3 4 5).')
parser.add_argument('-e', '--exclude-groups', nargs='*', type=int,
help='List of device group IDs to exclude from polling/discovery.')
parser.add_argument('-d', '--debug', action='store_true',
help='Enable debug output. WARNING, do not use this option unless you know what it does. This '
'generates a lot of very large files in TEMP dir.', default=False)
parser.add_argument('-t', '--test', action='store_true',
help='Do not spawn processes, just test DB connection, config, etc.', default=False)
parser.add_argument('--host', help='Process hostname wildcard.')
# Parse passed arguments
args = parser.parse_args()
# print(args)
# for compatibility with old passed argument with worker
try:
# poller-wrapper.py 16
workers = int(args.process)
workers_arg = False
process = 'poller'
print("WARNING! Using number of threads without command argument (-w or --workers) is deprecated! \n"
"Please replace your poller wrapper command line with:")
print("observium-wrapper poller -w %s" % workers)
except ValueError:
# observium-wrapper poller -w 16
process = args.process
workers = int(args.workers)
poller_id = int(args.poller_id)
instances_count = int(args.instances)
instance_number = int(args.number)
stats = args.stats
debug = args.debug
test = args.test
except ImportError:
# FIXME, do not use this compatibility, just end execution here
print("WARNING: missing the argparse python module:")
print("On Ubuntu: apt-get install libpython%s.%s-stdlib" % python_version)
print("On RHEL/CentOS: yum install python-argparse")
print("On Debian: apt-get install python-argparse")
print("Continuing with basic argument support.")
poller_id = -1
instances_count = 1
instance_number = 0
stats = False
debug = False
test = False
try:
# poller-wrapper.py 16
workers = int(sys.argv[1])
workers_arg = False
process = 'poller'
except ValueError:
# observium-wrapper poller -w 16
process = sys.argv[1]
if not process:
process = 'poller'
workers = 0
"""
Allowed process list, if not one of this exit with error
"""
process_list = ['poller', 'discovery', 'billing']
if process not in process_list:
print("ERROR: Incorrect process name '%s' passed, should be one of: %s" % (process, process_list))
sys.exit(2)
# entity name for log and messages
if process == 'billing':
entity = 'bill'
else:
entity = 'device'
# full process name, ie: 'observium-wrapper poller'
processname = scriptname + ' ' + process
"""
Fetch configuration details from the config_to_json.php script
"""
ob_install_dir = os.path.dirname(os.path.realpath(__file__))
config_file = ob_install_dir + '/config.php'
def get_config_data():
config_cmd = ['/usr/bin/env', 'php', '%s/config_to_json.php' % ob_install_dir]
# limit requested options only required (skip huge definitions)
config_options = ['db_user', 'db_pass', 'db_host', 'db_name', 'db_port', 'db_socket',
'install_dir', 'rrd_dir', 'temp_dir', 'log_dir', 'mib_dir',
'rrdcached', 'rrdtool', 'rrd',
'poller-wrapper', 'poller_id', 'poller_name']
config_cmd.append('-o')
config_cmd.append(','.join(config_options))
try:
proc = subprocess.Popen(config_cmd, stdout=subprocess.PIPE, stdin=subprocess.PIPE)
except:
print("ERROR: Could not execute: %s" % config_cmd)
sys.exit(2)
return proc.communicate()[0].decode('utf-8') # decode required in python3
try:
with open(config_file) as f:
pass
except IOError as e:
print("ERROR: Oh dear... %s does not seem readable" % config_file)
sys.exit(2)
try:
config = json.loads(get_config_data())
except:
print("ERROR: Could not load or parse observium configuration from config_to_json.php, are PATHs correct?")
sys.exit(2)
if test:
print(config)
db_username = config['db_user']
db_password = config['db_pass']
db_server = config['db_host']
db_dbname = config['db_name']
try:
db_port = int(config['db_port'])
except KeyError:
db_port = 3306
try:
db_socket = config['db_socket']
except KeyError:
db_socket = False
poller_path = config['install_dir'] + '/poller.php'
discovery_path = config['install_dir'] + '/discovery.php'
alerter_path = config['install_dir'] + '/alerter.php'
billing_path = config['install_dir'] + '/poll-billing.php'
# rrdcached & remote rrd
try:
rrdcached_address = config['rrdcached']
remote_rrd = rrdcached_address.find('/') < 0 # unix:/file or /file
except KeyError:
# rrdcached config not set, reset remote_rrd
remote_rrd = False
try:
temp_path = config['temp_dir']
except KeyError:
temp_path = '/tmp'
try:
rrd_path = config['rrd_dir']
except KeyError:
rrd_path = config['install_dir'] + '/rrd'
try:
log_path = config['log_dir'] + '/observium.log'
except KeyError:
log_path = config['install_dir'] + '/logs/observium.log'
# Amount of threads
if process == 'poller':
try:
# use config option if set
amount_of_workers = int(config['poller-wrapper']['threads'])
except KeyError:
amount_of_workers = 0
else:
amount_of_workers = 0
if workers > 0 and (workers_arg or amount_of_workers < 1):
# use amount threads passed as argument
# always prefer cmd arg
amount_of_workers = workers
if amount_of_workers < 1:
if process == 'poller':
max_workers = 256
else:
# set discovery workers limitation less than poller
max_workers = 8
try:
# use threads count based on cpus count
import multiprocessing
cpu_count = multiprocessing.cpu_count()
amount_of_workers = cpu_count * 2
# Limit maximum amount of worker based on CPU count
if amount_of_workers > max_workers:
print("WARNING: Very high CPU core count, %s threads limited to %s (detected: %s). For more threads"
" please use configuration option $config['poller-wrapper']['threads']"
" or pass as argument -w [THREADS]" % (process, max_workers, amount_of_workers))
amount_of_workers = max_workers
except (ImportError, NotImplementedError):
amount_of_workers = 8
print("WARNING: used default thread count of %s. To change this use configuration options"
" or pass as argument -w [THREADS]" % amount_of_workers)
if test:
cpu_count = 'unknown'
if 'cpu_count' not in locals():
try:
import multiprocessing
cpu_count = multiprocessing.cpu_count()
except (ImportError, NotImplementedError):
cpu_count = 'unknown'
print("Script: %s, Process: %s, Workers: %s, CPUs: %s, Instances: %s, InstanceID: %s" % (
scriptname, process, amount_of_workers, cpu_count, instances_count, instance_number))
print("Stats: %s, Remote RRD: %s, Debug: %s, Test: %s" % (stats, remote_rrd, debug, test))
print("Versions:\n Python - %s.%s.%s" % sys.version_info[:3])
print(" DB - %s" % db_version)
if os.path.isfile(alerter_path):
alerting = config['poller-wrapper']['alerter']
else:
alerting = False
if not stats:
try:
stats = bool(config['poller-wrapper']['stats'])
except KeyError:
pass
# max running and max LA
try:
max_running = int(config['poller-wrapper']['max_running'])
except:
max_running = 3
if max_running < 1:
max_running = 3
try:
max_la = float(config['poller-wrapper']['max_la'])
except:
max_la = 10
if max_la <= 0:
max_la = 10
# poller/discovery timeouts per device
# Notes:
# subprocess.run() was added in 3.5
# subprocess.check_call() can migrate to subprocess.run(..., check=True)
# timeout param was added to subprocess.check_call() in 3.3
timeout = 0
if entity == 'device' and python_version >= (3, 3):
# only for poller/discovery with python 3.3+
if int(args.timeout) > 0:
# timeout param passed from cmd line (can use any positive value)
timeout = int(args.timeout)
elif int(config['poller-wrapper'][process + '_timeout']) >= 90:
# hard limit can't be less than 90s (prevent use random numbers as timeout)
timeout = int(config['poller-wrapper'][process + '_timeout'])
# partitioned poller
poller_arg = False # pass poller id to requested script
if poller_id < 0:
# poller_id not passed from command line, use config or default
try:
poller_id = int(config['poller_id'])
except:
poller_id = 0
else:
# poller_arg = args.poller_id > 0
poller_arg = True
# poller_id = args.poller_id
if test:
print("Timeout: %ss" % timeout)
"""
Check mibs dir for stale .index files
"""
try:
import glob
# use this cleanup in discovery --host new (not in poller)
if process == 'discovery':
mib_indexes = glob.glob(config['mib_dir'] + '/.index')
mib_indexes += glob.glob(config['mib_dir'] + '/*/.index')
# mib_indexes += glob.glob('/var/lib/snmp/mib_indexes/*') # not permitted for wrapper process
# print(mib_indexes)
for mib_index in mib_indexes:
# print(mib_index)
try:
os.remove(mib_index)
except:
# break loop because not permitted to remove files in mibs dir
print("WARNING: .index files are found in mibs directories which can't be removed "
"(there aren't enough permissions)")
print(" see: http://www.observium.org/docs/faq/"
"#all-my-hosts-seem-down-to-observium-snmp-doesnt-seem-to-work-anymore")
logfile("WARNING: .index files are found in mibs directories which can't be removed "
"(there aren't enough permissions)")
break
except:
pass
# sys.exit(2)
real_duration = 0
per_device_duration = {}
devices_list = []
try:
# set db connection params
db_params = {'host': db_server,
'user': db_username,
'passwd': db_password,
'db': db_dbname,
'port': db_port}
if bool(db_socket):
db_params['unix_socket'] = db_socket
if "pymysql" in db_version:
# enable autocommit for pymysql lib
db_params['autocommit'] = True
db = MySQLdb.connect(**db_params)
if "MySQLdb" in db_version:
# enable autocommit for MySQLdb lib
db.autocommit(True)
cursor = db.cursor()
except:
print("ERROR: Could not connect to MySQL database!")
logfile("ERROR: Could not connect to MySQL database!")
sys.exit(2)
"""
This query specifically orders the results depending on the last_polled_timetaken variable
Because this way, we put the devices likely to be slow, in the top of the queue
thus greatening our chances of completing _all_ the work in exactly the time it takes to
poll the slowest device! cool stuff he
Additionally, if a hostname wildcard is passed, add it to the where clause. This is
important in cases where you have pollers distributed geographically and want to limit
pollers to polling hosts matching their geographic naming scheme.
"""
stop_script = False # trigger for stop execute script inside try
where_devices = "WHERE disabled != 1" # Filter disabled devices by default
# Use include device groups
if args.include_groups is not None:
# Fetch device_id from selected groups
query = "SELECT entity_id FROM group_table WHERE entity_type = 'device' AND entity_id > 0 AND group_id IN (%s)"
cursor.execute(query % ",".join(map(str, args.include_groups)))
include_devices = []
for row in cursor.fetchall():
include_devices.append(row[0])
# print(map(str, include_devices))
# print(query % ",".join(map(str, args.include_groups)))
where_devices += " AND device_id IN (%s)" % ",".join(map(str, include_devices))
# Use exclude device groups
if args.exclude_groups is not None:
# Fetch device_id from selected groups
query = "SELECT entity_id FROM group_table WHERE entity_type = 'device' AND entity_id > 0 AND group_id IN (%s)"
cursor.execute(query % ",".join(map(str, args.exclude_groups)))
exclude_devices = []
for row in cursor.fetchall():
exclude_devices.append(row[0])
# print(map(str, exclude_devices))
# print(query % ",".join(map(str, args.include_groups)))
where_devices += " AND device_id NOT IN (%s)" % ",".join(map(str, exclude_devices))
if poller_id > 0:
print(Colors.OKBLUE + 'INFO: This is poller_id (' + str(poller_id) + ') using poller-restricted devices list' +
Colors.ENDC)
else:
print(Colors.OKBLUE + 'This is the default poller. Will only poll devices with no specified poller set.' +
Colors.ENDC)
# Set default value of 0 for process tables and the like
poller_id = 0
# Always select devices by poller_id
where_devices += " AND poller_id = '" + str(poller_id) + "'"
if test:
print(where_devices)
if instances_count > 1 and instance_number >= 0 and (instance_number < instances_count):
# Use distributed wrapper
if process == 'billing':
# billing
query = """SELECT bill_id FROM (SELECT @rownum :=0) r,
(
SELECT @rownum := @rownum +1 AS rownum, bill_id
FROM bills
ORDER BY bill_id ASC
) temp
WHERE MOD(temp.rownum, %s) = %s""" % (instances_count, instance_number)
else:
# poller or discovery
query = """SELECT device_id FROM (SELECT @rownum :=0) r,
(
SELECT @rownum := @rownum +1 AS rownum, device_id
FROM devices
%s
ORDER BY device_id ASC
) temp
WHERE MOD(temp.rownum, %s) = %s""" % (where_devices, instances_count, instance_number)
if test:
print(query)
cursor.execute(query)
# Increase maximum allowed running wrapper processes by instances count
max_running *= instances_count
else:
# set instances count and number for single process wrapper
instances_count = 1
instance_number = 0
if process == 'billing':
# Normal billing poll
query = """SELECT bill_id
FROM bills"""
order = " ORDER BY bill_id ASC"
query = query + order
if test:
print(query)
cursor.execute(query)
else:
# Normal poller/discovery
query = """SELECT device_id
FROM devices
%s""" % (where_devices,)
if process == 'discovery':
order = " ORDER BY last_discovered_timetaken DESC"
else:
order = " ORDER BY last_polled_timetaken DESC"
# skip down devices in discovery
if process == 'discovery':
query = query + " AND status = '1'"
try:
# Query with hosts specified
host_wildcard = args.host.replace('*', '%')
# expand process name for do not calculate count inside main processes
processname = processname + ' --host ' + host_wildcard
# new devices discovery just pass ./discovery.php -h new, do not spawn processes
if process == 'discovery' and host_wildcard == 'new':
if debug:
command_options = '-d -h ' + host_wildcard
else:
command_options = '-q -h ' + host_wildcard
# append poller id arg when requested
if poller_arg:
command_options = '-p %s %s' % (poller_id, command_options)
print("INFO: starting %s for %s on poller id %s" % (process, host_wildcard, poller_id))
# print("/usr/bin/env php %s %s >> /dev/null 2>&1" % (discovery_path, command_options))
os.system("/usr/bin/env php %s %s >> /dev/null 2>&1" % (discovery_path, command_options))
print("INFO: finished %s for %s on poller id %s" % (process, host_wildcard, poller_id))
# sys.exit(0) # this not worked inside try,
# see: https://stackoverflow.com/questions/7709411/why-finally-block-is-executing-after-calling-sys-exit0-in-except-block
stop_script = True
# wc_query = query + " AND (last_discovered IS NULL OR last_discovered = '0000-00-00 00:00:00' OR force_discovery = '1')"
# cursor.execute(wc_query)
else:
wc_query = query + " AND hostname LIKE %s " + order
cursor.execute(wc_query, (host_wildcard,))
if test:
print(wc_query)
except:
# Query without hosts specified
query = query + order
if test:
print(query)
cursor.execute(query)
# stop script execute after try
if stop_script:
sys.exit(0)
devices = cursor.fetchall()
for row in devices:
devices_list.append(int(row[0]))
if test:
print(devices_list)
# sys.exit(2)
"""
Get current wrapper process info and remove stale db entries
"""
pid = os.getpid()
ppid = os.getppid()
uid = os.getuid()
la = os.getloadavg()
try:
command = subprocess.check_output('ps -ww -o args -p %s' % pid, shell=True, universal_newlines=True).splitlines()[1]
except:
command = scriptname
"""
Search if same poller wrapper proccesses running
Protect from race condition
"""
ps_count = 0
p_query = """SELECT `process_id`, `process_pid`, `process_ppid`, `process_uid`, `process_command`, `process_start`
FROM `observium_processes`
WHERE `process_name` = %s AND `poller_id` = %s"""
try:
cursor.execute(p_query, (processname, poller_id))
for row in cursor.fetchall():
# print(row)
test_running = False
test_id, test_pid, test_ppid, test_uid, test_command, test_start = row
try:
test_ps = subprocess.check_output('ps -ww -o ppid,uid,args -p %s' % test_pid, shell=True,
universal_newlines=True).splitlines()[1]
# print(test_ps)
test_ps = test_ps.split(None, 2)
# print(test_ps)
# print("PPID: %s, %s" % (test_ppid, int(test_ps[0])))
# print(" UID: %s, %s" % (test_uid, int(test_ps[1])))
# print("name: %s, %s" % (processname, test_ps[2]))
test_running = (test_ppid == int(test_ps[0])) and (test_uid == int(test_ps[1])) and (
scriptname in test_ps[2])
except:
# not exist pid
pass
# print("Test: %s" % (test_running))
if not test_running:
# process not exist, remove stale db entry
try:
cursor.execute("""DELETE FROM `observium_processes` WHERE `process_id` = %s""", (test_id,))
# db.commit()
# print("Removed stale DB entry %s" % test_id)
except:
pass
else:
ps_count += 1
# print("Count: %s" % (ps_count))
except:
try:
# FIXME. Remove this compatibility, always use from db process counts!
ps_list = subprocess.check_output('ps ax | grep %s | grep -v grep' % scriptname, shell=True,
universal_newlines=True)
# divide by 2 because cron starts 2 processes (/bin/sh and main process)
ps_count = len(ps_list.splitlines()) / 2
ps_count -= 1 # decrease current process
except:
# Skip searching, something wrong
pass
# This prevents race and too high LA on server.
# Default is 4 processes and 10 load average.
# More than 4 already running poller-wrapper it's big trouble!
if ps_count > max_running and la[1] >= max_la:
print("URGENT: %s not started because already running %s processes, load average (5min) %.2f" % (
processname, ps_count, la[1]))
logfile("URGENT: %s not started because already running %s processes, load average (5min) %.2f" % (
processname, ps_count, la[1]))
sys.exit(2)
# Increase count by current wrapper process
ps_count += 1
# write into db current process info
p_query = """INSERT INTO `observium_processes` (`process_pid`,`process_ppid`,`process_name`,`process_uid`,`process_command`,`process_start`,`device_id`,`poller_id`)
VALUES (%s,%s,%s,%s,%s,%s,'0',%s)"""
try:
cursor.execute(p_query, (pid, ppid, processname, uid, command, s_time, poller_id))
process_id = db.insert_id()
# db.commit()
except:
process_id = None
pass
if test:
print("Already running %s processes: %s, load average (5min) %.2f" % (processname, ps_count, la[1]))
# print(process_id)
# time.sleep(30) # delays for 30 seconds
if process_id is not None:
p_query = """DELETE FROM `observium_processes` WHERE `process_id` = %s"""
cursor.execute(p_query, (process_id,))
# db.commit()
sys.exit(2)
# Open dev null handle
FNULL = open(os.devnull, 'w')
"""
Since ./discovery -h all additionally do db schema update,
also do this here before wrap discovery processes.
Check if this is first process instance and --host parameter not passed
"""
if process == 'discovery' and instance_number == 0 and 'host_wildcard' not in globals():
# print("/usr/bin/env php %s -q -u >> /dev/null 2>&1" % (discovery_path))
if debug:
command_options = '-d -u -a'
else:
command_options = '-q -u -a'
print("INFO: starting discovery.php for update")
os.system("/usr/bin/env php %s %s >> /dev/null 2>&1" % (discovery_path, command_options))
print("INFO: finished discovery.php for update")
def update_wrapper_times(rrd_file, count, runtime, workers):
"""
Create/update poller stat times
"""
if not remote_rrd:
rrd_file = rrd_path + '/' + rrd_file
# always create rrd (with no-overwrite) for remote rrdcached
if remote_rrd or not os.path.isfile(rrd_file):
# Create RRD
rrd_dst = ':GAUGE:' + str(config['rrd']['step'] * 2) + ':0:U'
cmd_create = config['rrdtool'] + ' create ' + rrd_file + ' DS:devices' + rrd_dst + ' DS:totaltime'
cmd_create += rrd_dst +' DS:threads' + rrd_dst
cmd_create += ' --step ' + str(config['rrd']['step']) + ' ' + ' '.join(config['rrd']['rra'].split())
if remote_rrd:
# --no-overwrite available since 1.4.3
cmd_create += ' --no-overwrite --daemon ' + rrdcached_address + ' >/dev/null 2>&1'
else:
logfile(cmd_create)
os.system(cmd_create)
if debug:
print("DEBUG: " + cmd_create)
cmd_update = config['rrdtool'] + ' update ' + rrd_file + ' N:%s:%s:%s' % (count, runtime, workers)
if remote_rrd:
cmd_update += ' --daemon ' + rrdcached_address
os.system(cmd_update)
if debug:
print("DEBUG: " + cmd_update)
def update_wrapper_count(rrd_file, count):
"""
Create/update poller wrapper count
"""
if not remote_rrd:
rrd_file = rrd_path + '/' + rrd_file
if remote_rrd or not os.path.isfile(rrd_file):
# Create RRD
rrd_dst = ':GAUGE:' + str(config['rrd']['step'] * 2) + ':0:U'
cmd_create = config['rrdtool'] + ' create ' + rrd_file + ' DS:wrapper_count' + rrd_dst
cmd_create += ' --step ' + str(config['rrd']['step']) + ' ' + ' '.join(config['rrd']['rra'].split())
if remote_rrd:
# --no-overwrite available since 1.4.3
cmd_create += ' --no-overwrite --daemon ' + rrdcached_address + ' >/dev/null 2>&1'
else:
logfile(cmd_create)
os.system(cmd_create)
if debug:
print("DEBUG: " + cmd_create)
cmd_update = config['rrdtool'] + ' update ' + rrd_file + ' N:%s' % count
if remote_rrd:
cmd_update += ' --daemon ' + rrdcached_address
os.system(cmd_update)
if debug:
print("DEBUG: " + cmd_update)
def printworker():
"""
A seperate queue and a single worker for printing information to the screen prevents
the good old joke:
Some people, when confronted with a problem, think,
"I know, I'll use threads," and then two they hav erpoblesms.
"""
while True:
worker_id, device_id, elapsed_time = print_queue.get()
global real_duration
global per_device_duration
real_duration += elapsed_time
per_device_duration[device_id] = elapsed_time
if elapsed_time < 300:
print("INFO: worker %s finished %s %s in %s seconds" % (worker_id, entity, device_id, elapsed_time))
else:
print("WARNING: worker %s finished %s %s in %s seconds" % (worker_id, entity, device_id, elapsed_time))
print_queue.task_done()
def process_worker():
"""
This class will fork off single instances of the poller.php process, record
how long it takes, and push the resulting reports to the printer queue
"""
command_paths = {
'poller': poller_path,
'discovery': discovery_path,
'billing': billing_path
}
process_path = command_paths[process]
command_list = []
command_out = FNULL # set non-debug output to /dev/null
if process == 'billing':
# billing use -b [bill_id]
command_list.append('-b')
else:
# poller/discovery us -h [device_id]
if poller_arg:
# add poller id arg
command_list.append('-p')
command_list.append(poller_id)
command_list.append('-h')
if debug:
command_list.insert(0, '-d')
else:
command_list.insert(0, '-q')
while True:
device_id = process_queue.get()
try:
start_time = time.time()
command_args = ['/usr/bin/env', 'php', process_path]
command_args.extend(command_list)
command_args.append(device_id)
if debug:
command_out = open(temp_path + '/observium_' + process + '_' + str(device_id) + '.debug', 'w')
# print(command_args) #debug
if timeout > 0:
# append timeout param for python 3.3+ and when option set
try:
subprocess.check_call(map(str, command_args),
stdout=command_out, stderr=subprocess.STDOUT, timeout=timeout)
except subprocess.TimeoutExpired:
print("WARNING: %s for device_id %s ran too long, stopped by timeout." % (process, device_id))
logfile("WARNING: %s for device_id %s ran too long, stopped by timeout." % (process, device_id))
else:
subprocess.check_call(map(str, command_args), stdout=command_out, stderr=subprocess.STDOUT)
if debug:
command_out.close()
# additional alerter process only after poller process
if process == 'poller' and alerting:
print("INFO: starting alerter.php for %s" % device_id)
command_args = ['/usr/bin/env', 'php', alerter_path]
command_args.extend(command_list)
command_args.append(device_id)
if debug:
command_out = temp_path + '/observium_alerter_' + str(device_id) + '.debug'
# print(command_args) #debug
subprocess.check_call(map(str, command_args), stdout=FNULL, stderr=subprocess.STDOUT)
print("INFO: finished alerter.php for %s" % device_id)
elapsed_time = int(time.time() - start_time)
print_queue.put([threading.current_thread().name, device_id, elapsed_time])
except (KeyboardInterrupt, SystemExit):
raise
except:
pass
process_queue.task_done()
process_queue = Queue.Queue()
print_queue = Queue.Queue()
print("INFO: starting the %s at %s with %s threads" % (process, time.strftime("%Y/%m/%d %H:%M:%S"), amount_of_workers))
if debug:
print("WARNING: DEBUG enabled, each %s %s store output to %s/observium_%s_id.debug (where id is %s_id)" % (
entity, process, temp_path, process, entity))
for device_id in devices_list:
process_queue.put(device_id)
for i in range(amount_of_workers):
t = threading.Thread(target=process_worker)
t.daemon = True
t.start()
p = threading.Thread(target=printworker)
p.daemon = True
p.start()
try:
process_queue.join()
print_queue.join()
except (KeyboardInterrupt, SystemExit):
raise
total_time = time.time() - s_time
la = os.getloadavg() # get new load average after polling
msg = "processed %s %ss in %.2f seconds with %s threads, load average (5min) %.2f" % \
(len(devices_list), entity, total_time, amount_of_workers, la[1])
print("INFO: %s %s\n" % (processname, msg))
logfile(msg)
show_stopper = False
exit_code = 0
if process != 'discovery' and total_time > 300:
recommend = int(total_time / 300.0 * amount_of_workers + 1)
print("WARNING: the process took more than 5 minutes to finish, you need faster hardware or more threads")
print("INFO: in sequential style processing the elapsed time would have been: %s seconds" % real_duration)
for device in per_device_duration:
if per_device_duration[device] > 300:
print("WARNING: device %s is taking too long: %s seconds" % (device, per_device_duration[device]))
show_stopper = True
if show_stopper:
print("ERROR: Some devices are taking more than 300 seconds, the script cannot recommend you what to do.")
else:
print("WARNING: Consider setting a minimum of %d threads. (This does not constitute professional advice!)" %
recommend)
exit_code = 1
"""
Get and update poller wrapper stats
"""
if process == 'poller' and stats:
import socket
localhost = socket.getfqdn()
if localhost.find('.') == 0:
localhost_t = subprocess.check_output(["/bin/hostname", "-f"], universal_newlines=True).strip()
if localhost_t.find('.') > 0:
localhost = localhost_t
poller_stats = {
'instances_count': instances_count,
'wrapper_count': ps_count
}
instance_stats = {
'hostname': localhost,
'instances_count': instances_count, # this is need for check in calculate average
'workers_count': amount_of_workers,
'devices_count': len(devices_list),
'devices_ids': devices_list,
'last_starttime': '%.3f' % s_time,
'last_runtime': '%.3f' % total_time,
'last_status': exit_code,
# 'last_message': msg
}
instances_clean = []
instances_total_time = total_time
instances_total_devices = len(devices_list)
i = 1
query = """SELECT attrib_type, attrib_value FROM observium_attribs WHERE attrib_type LIKE 'poller_wrapper_instance_%'"""
cursor.execute(query)
for row in cursor.fetchall():
# print(row)
n = int(row[0].replace('poller_wrapper_instance_', '')) # get stat for instance number n
if n >= instances_count:
# seems as old stats, instances count reduced
instances_clean.append(row[0])
elif n == instance_number:
# current instance, skip
pass
else:
instance = json.loads(row[1])
if int(instance['instances_count']) == instances_count:
# print(instance)
i += 1
instances_total_time += float(instance['last_runtime'])
instances_total_devices += int(instance['devices_count'])
else:
instances_clean.append(row[0])
instances_average_time = instances_total_time / i
poller_stats['total_devices_count'] = instances_total_devices
poller_stats['total_runtime'] = '%.3f' % instances_total_time
poller_stats['average_runtime'] = '%.3f' % instances_average_time
# print(poller_stats)
# print(instance_stats)
# print(instances_clean)
# write instance and total stats into db
cursor.execute("SELECT EXISTS (SELECT 1 FROM `observium_attribs` WHERE `attrib_type` = %s)",
('poller_wrapper_instance_' + str(instance_number),))
row = cursor.fetchone()
if int(row[0]) > 0:
cursor.execute("UPDATE `observium_attribs` SET `attrib_value` = %s WHERE `attrib_type` = %s",
(json.dumps(instance_stats), 'poller_wrapper_instance_' + str(instance_number)))
# db.commit()
# print("Number of rows updated: %d" % cursor.rowcount)
else:
cursor.execute("INSERT INTO `observium_attribs` (`attrib_value`,`attrib_type`) VALUES (%s,%s)",
(json.dumps(instance_stats), 'poller_wrapper_instance_' + str(instance_number)))
# db.commit()
cursor.execute("SELECT EXISTS (SELECT 1 FROM `observium_attribs` WHERE `attrib_type` = %s)",
('poller_wrapper_stats',))
row = cursor.fetchone()
if int(row[0]) > 0:
cursor.execute("UPDATE `observium_attribs` SET `attrib_value` = %s WHERE `attrib_type` = %s",
(json.dumps(poller_stats), 'poller_wrapper_stats'))
# db.commit()
# print("Number of rows updated: %d" % cursor.rowcount)
else:
cursor.execute("INSERT INTO `observium_attribs` (`attrib_value`,`attrib_type`) VALUES (%s,%s)",
(json.dumps(poller_stats), 'poller_wrapper_stats'))
# db.commit()
# clean stale instance stats
for row in instances_clean:
cursor.execute("DELETE FROM `observium_attribs` WHERE `attrib_type` = %s", (row,))
# db.commit()
# Write poller statistics to RRD
if poller_id > 0:
# partitioned poller wrapper
rrd_name_base = 'poller-wrapper-id' + str(poller_id)
else:
# default poller wrapper
rrd_name_base = 'poller-wrapper'
rrd_path_total = rrd_name_base + '.rrd'
rrd_path_count = rrd_name_base + '_count.rrd'
update_wrapper_times(rrd_path_total, poller_stats['total_devices_count'], poller_stats['average_runtime'],
amount_of_workers)
update_wrapper_count(rrd_path_count, poller_stats['wrapper_count'])
# Additionally per instance statistics
if instances_count > 1:
rrd_path_instance = rrd_name_base + '_' + str(instances_count) + '_' + str(instance_number) + '.rrd'
update_wrapper_times(rrd_path_instance, instance_stats['devices_count'], instance_stats['last_runtime'],
amount_of_workers)
# Remove process info from DB
p_query = """DELETE FROM `observium_processes` WHERE `process_id` = %s"""
try:
cursor.execute(p_query, (process_id,))
# db.commit()
except:
pass
# Close opened handles
db.close()
FNULL.close()
# Return exit code
sys.exit(exit_code)
# EOF