zulip/scripts/nagios/check-rabbitmq-consumers

#!/usr/bin/env python

from __future__ import print_function
import sys
import time
import optparse
from collections import defaultdict
import os
import subprocess

states = {
    0: "OK",
    1: "WARNING",
    2: "CRITICAL",
    3: "UNKNOWN"
}

if 'USER' in os.environ and not os.environ['USER'] in ['root', 'rabbitmq']:
    print("This script must be run as the root or rabbitmq user")


usage = """Usage: check-rabbitmq-consumers --queue=[queue-name] --min-threshold=[min-threshold]"""

parser = optparse.OptionParser(usage=usage)
parser.add_option('--min-threshold',
                  dest='min_count',
                  type="int",
                  default=1,
                  action='store')

(options, args) = parser.parse_args()

output = subprocess.check_output(['/usr/sbin/rabbitmqctl', 'list_consumers'],
                                 universal_newlines=True)

consumers = defaultdict(int) # type: Dict[str, int]

worker_queues = {'error_reports',
                 'user_presence',
                 'digest_emails',
                 'slow_queries',
                 'missedmessage_mobile_notifications',
                 'feedback_messages',
                 'signups',
                 'notify_tornado',
                 'message_sender',
                 'missedmessage_emails',
                 'email_mirror',
                 'user_activity_interval',
                 'invites',
                 'user_activity'}

for queue_name in worker_queues:
    consumers[queue_name] = 0

for line in output.split('\n'):
    parts = line.split('\t')
    if len(parts) >= 2:
        consumers[parts[0]] += 1

now = int(time.time())

for queue_name in consumers.keys():
    state_file_path = "/var/lib/nagios_state/check-rabbitmq-consumers-" + queue_name
    state_file_tmp = state_file_path + "-tmp"

    if consumers[queue_name] < options.min_count:
        status = 2
    else:
        status = 0
    with open(state_file_tmp, "w") as f:
        f.write("%s|%s|%s|queue %s has %s consumers, needs %s\n" % (
            now, status, states[status], queue_name,
            consumers[queue_name], options.min_count))
    subprocess.check_call(["mv", state_file_tmp, state_file_path])