[uio-zabbix.git] / zabbix_get_rabbitmq_stats.py

#!/usr/bin/python2
#
# Authors:
# rafael@E-MC2.NET / https://e-mc2.net/
#
# Copyright (c) 2018 USIT-University of Oslo
#
# zabbix_get_rabbitmq_stats.py is free software: you can redistribute
# it and/or modify it under the terms of the GNU General Public
# License as published by the Free Software Foundation, either version
# 3 of the License, or (at your option) any later version.
#
# zabbix_get_rabbitmq_stats.py is distributed in the hope that it will
# be useful, but WITHOUT ANY WARRANTY; without even the implied
# warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
# See the GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with sms_send.  If not, see <http://www.gnu.org/licenses/>.
#

import requests
import json
import sys
import os
import tempfile
import socket
import subprocess

# Zabbix proxy
zabbix_proxy = ['zabbix-proxy-prod03.uio.no','zabbix-proxy-prod04.uio.no']

# Path to zabbix_sender
zabbix_sender = '/usr/bin/zabbix_sender'

# RabbitMQ server domain
domain = ".uio.no"

# Authentication file with username / password used to access RabbitMQ
# API.
auth_file = "/var/lib/zabbix/.rabbitmq_auth"


# ############################################
# get_auth_data()
# ############################################

def get_auth_data():
    
    try:

        username = "none"
        password = "none"

        if os.path.isfile(auth_file):

            with open(auth_file, 'r') as f:
                for line in f:
                    (username, password) = line.split('::')
                    
                password = password.replace('\n', '')

        return username,password
            
    except Exception as e:
        raise Exception("[ERROR]: %s\n" % e)


# ############################################
# get_rabbitmq_data()
# ############################################

def get_rabbitmq_data(component):
    
    try:
        (username,password) = get_auth_data()
        request_data = requests.get("http://127.0.0.1:15672/api/" + component ,auth=(username,password))
                
        if request_data.status_code != 200:
            raise Exception("[ERROR]: Problems connecting to rabbitMQ stats API\n")
            
        data = request_data.json()

        return data
            
    except Exception as e:
        raise Exception("[ERROR]: %s\n" % e)
 

# ############################################
# generate_zabbix_autodiscovery()
# ############################################

def generate_zabbix_sender_data(data,component):

    try:
        result = ""
        hostname = socket.gethostname()

        if component == "queues":

            for queue in data:
                if hostname.replace(domain,'') in queue['node']:
            
                    result +=  hostname + " queue.consumers[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['consumers']) + "\n"

                    # consumer_utilisation is defined as none when the
                    # queue is not in use.
                    
                    if queue['consumer_utilisation'] == None:
                        queue['consumer_utilisation'] = 0
                        
                    result +=  hostname + " queue.consumer.utilisation[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['consumer_utilisation']) + "\n"
                    result +=  hostname + " queue.memory[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['memory']) + "\n"
                    result +=  hostname + " queue.messages[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['messages']) + "\n"
                    result +=  hostname + " queue.message.bytes[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_bytes']) + "\n"

                    # message_stats is not defined when the queue is
                    # not in use.
                    
                    if 'message_stats' in queue:
                    
                        result +=  hostname + " queue.messages.deliver.get[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_stats']['deliver_get']) + "\n"
                        result +=  hostname + " queue.messages.publish[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_stats']['publish']) + "\n"
                    
                    result +=  hostname + " queue.messages.ram[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['messages_ram']) + "\n"
                    result +=  hostname + " queue.node[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['node']) + "\n"
                    result +=  hostname + " queue.state[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['state']) + "\n"
              
        elif component == "nodes":

            for node in data:
                if hostname.replace(domain,'') in node['name']:
                
                    result +=  hostname + " node.disk.free " + str(node['disk_free']) + "\n"
                    result +=  hostname + " node.disk.free.alarm " + str(node['disk_free_alarm']) + "\n"
                    result +=  hostname + " node.disk.free.limit " + str(node['disk_free_limit']) + "\n"
                    result +=  hostname + " node.proc.total " + str(node['proc_total']) + "\n"
                    result +=  hostname + " node.proc.used " + str(node['proc_used']) + "\n"
                    result +=  hostname + " node.run.queue " + str(node['run_queue']) + "\n"
                    result +=  hostname + " node.fd.total " + str(node['fd_total']) + "\n"
                    result +=  hostname + " node.fd.used " + str(node['fd_used']) + "\n"
                    result +=  hostname + " node.io.read.bytes " + str(node['io_read_bytes']) + "\n"
                    result +=  hostname + " node.io.read.count " + str(node['io_read_count']) + "\n"
                    result +=  hostname + " node.io.seek.count " + str(node['io_seek_count']) + "\n"
                    result +=  hostname + " node.io.sync.count " + str(node['io_sync_count']) + "\n"
                    result +=  hostname + " node.io.write.bytes " + str(node['io_write_bytes']) + "\n"
                    result +=  hostname + " node.io.write.count " + str(node['io_write_count']) + "\n"
                    result +=  hostname + " node.mem.used " + str(node['mem_used']) + "\n"
                    result +=  hostname + " node.mem.alarm " + str(node['mem_alarm']) + "\n"
                    result +=  hostname + " node.mem.limit " + str(node['mem_limit']) + "\n"
                    result +=  hostname + " node.queue.index.journal.write.count " + str(node['queue_index_journal_write_count']) + "\n"
                    result +=  hostname + " node.sockets.total " + str(node['sockets_total']) + "\n"
                    result +=  hostname + " node.sockets.used " + str(node['sockets_used']) + "\n"
                    result +=  hostname + " node.uptime " + str(node['uptime']) + "\n"
                  
        elif component == "overview":
        
            result +=  hostname + " cluster.name " + str(data['cluster_name']) + "\n"
            result +=  hostname + " cluster.rabbitmq.version " + str(data['rabbitmq_version']) + "\n"
            result +=  hostname + " cluster.erlang.version " + str(data['erlang_version']) + "\n" 
            result +=  hostname + " cluster.messages " + str(data['queue_totals']['messages']) + "\n" 
            result +=  hostname + " cluster.messages.deliver.get " + str(data['message_stats']['deliver_get']) + "\n" 
            result +=  hostname + " cluster.messages.publish " + str(data['message_stats']['publish']) + "\n" 
            result +=  hostname + " cluster.total.channels " + str(data['object_totals']['channels']) + "\n" 
            result +=  hostname + " cluster.total.connections " + str(data['object_totals']['connections']) + "\n" 
            result +=  hostname + " cluster.total.consumers " + str(data['object_totals']['consumers']) + "\n" 
            result +=  hostname + " cluster.total.exchanges " + str(data['object_totals']['exchanges']) + "\n" 
            result +=  hostname + " cluster.total.queues " + str(data['object_totals']['queues']) + "\n" 

        return result
                           
    except Exception as e:
      raise Exception("[ERROR]: %s\n" % e)


if __name__ == '__main__':

    try:

        #
        # We send data about queues, nodes and cluster to Zabbix
        #
        
        for component in ['queues','nodes','overview']:

            # Temp file with full json output
            tmp_stat_file = tempfile.NamedTemporaryFile(delete=False,dir='/tmp')

            data = get_rabbitmq_data(component)
            result = generate_zabbix_sender_data(data,component)

            #
            # We create a file with the data that zabbix_sender will
            # send in a bulk execution.
            #

            with open(tmp_stat_file.name,'w') as f: 
                f.write(result)
               
            #
            # The monitoring of this host can be done by any of the
            # zabbix proxies defined in zabbix_proxy[]. We try all of
            # them until one of them accepts our data
            #
        
            for proxy in zabbix_proxy:

                command = zabbix_sender + ' -z ' + proxy + ' -i ' + tmp_stat_file.name + ' > /dev/null 2>&1'
                
                proc = subprocess.Popen([command],shell=True)
                proc.wait()
                
                if proc.returncode == 0:
                    break

            # Delete temp file with zabbix_sender data
            os.remove(tmp_stat_file.name)

        # Return value 0 = execution OK
        print("0")

    except Exception, e:
        print("1")
        sys.exit(1)
Commit	Line	Data
ad51e17e	1	#!/usr/bin/python2
8861bf39 RM	2	#
	3	# Authors:
	4	# rafael@E-MC2.NET / https://e-mc2.net/
	5	#
	6	# Copyright (c) 2018 USIT-University of Oslo
	7	#
	8	# zabbix_get_rabbitmq_stats.py is free software: you can redistribute
	9	# it and/or modify it under the terms of the GNU General Public
	10	# License as published by the Free Software Foundation, either version
	11	# 3 of the License, or (at your option) any later version.
	12	#
	13	# zabbix_get_rabbitmq_stats.py is distributed in the hope that it will
	14	# be useful, but WITHOUT ANY WARRANTY; without even the implied
	15	# warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
	16	# See the GNU General Public License for more details.
	17	#
	18	# You should have received a copy of the GNU General Public License
	19	# along with sms_send. If not, see <http://www.gnu.org/licenses/>.
	20	#
	21
	22	import requests
	23	import json
	24	import sys
	25	import os
	26	import tempfile
	27	import socket
	28	import subprocess
	29
	30	# Zabbix proxy
	31	zabbix_proxy = ['zabbix-proxy-prod03.uio.no','zabbix-proxy-prod04.uio.no']
	32
	33	# Path to zabbix_sender
	34	zabbix_sender = '/usr/bin/zabbix_sender'
	35
	36	# RabbitMQ server domain
	37	domain = ".uio.no"
	38
	39	# Authentication file with username / password used to access RabbitMQ
	40	# API.
	41	auth_file = "/var/lib/zabbix/.rabbitmq_auth"
	42
	43
	44	# ############################################
	45	# get_auth_data()
	46	# ############################################
	47
	48	def get_auth_data():
	49
	50	try:
	51
	52	username = "none"
	53	password = "none"
	54
	55	if os.path.isfile(auth_file):
	56
	57	with open(auth_file, 'r') as f:
	58	for line in f:
	59	(username, password) = line.split('::')
	60
	61	password = password.replace('\n', '')
	62
	63	return username,password
	64
	65	except Exception as e:
66	raise Exception("[ERROR]: %s\n" % e)
67
68
69	# ############################################
70	# get_rabbitmq_data()
71	# ############################################
72
73	def get_rabbitmq_data(component):
74
75	try:
76	(username,password) = get_auth_data()
77	request_data = requests.get("http://127.0.0.1:15672/api/" + component ,auth=(username,password))
78
79	if request_data.status_code != 200:
d2217524	80	raise Exception("[ERROR]: Problems connecting to rabbitMQ stats API\n")
8861bf39 RM	81
	82	data = request_data.json()
	83
	84	return data
	85
	86	except Exception as e:
	87	raise Exception("[ERROR]: %s\n" % e)
	88
	89
	90	# ############################################
	91	# generate_zabbix_autodiscovery()
	92	# ############################################
	93
	94	def generate_zabbix_sender_data(data,component):
	95
	96	try:
	97	result = ""
	98	hostname = socket.gethostname()
	99
	100	if component == "queues":
	101
	102	for queue in data:
	103	if hostname.replace(domain,'') in queue['node']:
	104
	105	result += hostname + " queue.consumers[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['consumers']) + "\n"
	106
	107	# consumer_utilisation is defined as none when the
	108	# queue is not in use.
	109
	110	if queue['consumer_utilisation'] == None:
	111	queue['consumer_utilisation'] = 0
	112
	113	result += hostname + " queue.consumer.utilisation[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['consumer_utilisation']) + "\n"
	114	result += hostname + " queue.memory[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['memory']) + "\n"
	115	result += hostname + " queue.messages[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['messages']) + "\n"
	116	result += hostname + " queue.message.bytes[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_bytes']) + "\n"
	117
	118	# message_stats is not defined when the queue is
	119	# not in use.
	120
	121	if 'message_stats' in queue:
	122
	123	result += hostname + " queue.messages.deliver.get[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_stats']['deliver_get']) + "\n"
	124	result += hostname + " queue.messages.publish[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['message_stats']['publish']) + "\n"
	125
	126	result += hostname + " queue.messages.ram[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['messages_ram']) + "\n"
	127	result += hostname + " queue.node[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['node']) + "\n"
	128	result += hostname + " queue.state[" + queue['vhost'] + "," + queue['name'] + "] " + str(queue['state']) + "\n"
	129
	130	elif component == "nodes":
	131
	132	for node in data:
	133	if hostname.replace(domain,'') in node['name']:
	134
	135	result += hostname + " node.disk.free " + str(node['disk_free']) + "\n"
	136	result += hostname + " node.disk.free.alarm " + str(node['disk_free_alarm']) + "\n"
	137	result += hostname + " node.disk.free.limit " + str(node['disk_free_limit']) + "\n"
	138	result += hostname + " node.proc.total " + str(node['proc_total']) + "\n"
	139	result += hostname + " node.proc.used " + str(node['proc_used']) + "\n"
	140	result += hostname + " node.run.queue " + str(node['run_queue']) + "\n"
	141	result += hostname + " node.fd.total " + str(node['fd_total']) + "\n"
	142	result += hostname + " node.fd.used " + str(node['fd_used']) + "\n"
	143	result += hostname + " node.io.read.bytes " + str(node['io_read_bytes']) + "\n"
	144	result += hostname + " node.io.read.count " + str(node['io_read_count']) + "\n"
145	result += hostname + " node.io.seek.count " + str(node['io_seek_count']) + "\n"
146	result += hostname + " node.io.sync.count " + str(node['io_sync_count']) + "\n"
147	result += hostname + " node.io.write.bytes " + str(node['io_write_bytes']) + "\n"
148	result += hostname + " node.io.write.count " + str(node['io_write_count']) + "\n"
149	result += hostname + " node.mem.used " + str(node['mem_used']) + "\n"
150	result += hostname + " node.mem.alarm " + str(node['mem_alarm']) + "\n"
151	result += hostname + " node.mem.limit " + str(node['mem_limit']) + "\n"
152	result += hostname + " node.queue.index.journal.write.count " + str(node['queue_index_journal_write_count']) + "\n"
153	result += hostname + " node.sockets.total " + str(node['sockets_total']) + "\n"
154	result += hostname + " node.sockets.used " + str(node['sockets_used']) + "\n"
155	result += hostname + " node.uptime " + str(node['uptime']) + "\n"
156
157	elif component == "overview":
158
159	result += hostname + " cluster.name " + str(data['cluster_name']) + "\n"
160	result += hostname + " cluster.rabbitmq.version " + str(data['rabbitmq_version']) + "\n"
161	result += hostname + " cluster.erlang.version " + str(data['erlang_version']) + "\n"
162	result += hostname + " cluster.messages " + str(data['queue_totals']['messages']) + "\n"
163	result += hostname + " cluster.messages.deliver.get " + str(data['message_stats']['deliver_get']) + "\n"
164	result += hostname + " cluster.messages.publish " + str(data['message_stats']['publish']) + "\n"
165	result += hostname + " cluster.total.channels " + str(data['object_totals']['channels']) + "\n"
166	result += hostname + " cluster.total.connections " + str(data['object_totals']['connections']) + "\n"
167	result += hostname + " cluster.total.consumers " + str(data['object_totals']['consumers']) + "\n"
168	result += hostname + " cluster.total.exchanges " + str(data['object_totals']['exchanges']) + "\n"
169	result += hostname + " cluster.total.queues " + str(data['object_totals']['queues']) + "\n"
170
171	return result
172
173	except Exception as e:
174	raise Exception("[ERROR]: %s\n" % e)
175
176
177	if __name__ == '__main__':
178
179	try:
180
d2217524 RM	181	#
	182	# We send data about queues, nodes and cluster to Zabbix
	183	#
	184
8861bf39 RM	185	for component in ['queues','nodes','overview']:
	186
	187	# Temp file with full json output
	188	tmp_stat_file = tempfile.NamedTemporaryFile(delete=False,dir='/tmp')
	189
	190	data = get_rabbitmq_data(component)
	191	result = generate_zabbix_sender_data(data,component)
	192
	193	#
	194	# We create a file with the data that zabbix_sender will
	195	# send in a bulk execution.
	196	#
	197
	198	with open(tmp_stat_file.name,'w') as f:
	199	f.write(result)
	200
	201	#
	202	# The monitoring of this host can be done by any of the
d2217524	203	# zabbix proxies defined in zabbix_proxy[]. We try all of
8861bf39 RM	204	# them until one of them accepts our data
	205	#
	206
	207	for proxy in zabbix_proxy:
	208
	209	command = zabbix_sender + ' -z ' + proxy + ' -i ' + tmp_stat_file.name + ' > /dev/null 2>&1'
	210
d2217524	211	proc = subprocess.Popen([command],shell=True)
8861bf39 RM	212	proc.wait()
	213
	214	if proc.returncode == 0:
	215	break
	216
	217	# Delete temp file with zabbix_sender data
	218	os.remove(tmp_stat_file.name)
	219
	220	# Return value 0 = execution OK
	221	print("0")
	222
	223	except Exception, e:
	224	print("1")
	225	sys.exit(1)