zulip/scripts/nagios/check-rabbitmq-consumers

#!/usr/bin/env python3
import argparse
import os
import subprocess
import sys
import time
from collections import defaultdict
from typing import Dict

ZULIP_PATH = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
sys.path.append(ZULIP_PATH)
from scripts.lib.check_rabbitmq_queue import normal_queues
from scripts.lib.zulip_tools import get_config_file, get_tornado_ports

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

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


parser = argparse.ArgumentParser()
parser.parse_args()

config_file = get_config_file()
TORNADO_PROCESSES = len(get_tornado_ports(config_file))

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

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

queues = {
    *normal_queues,
    # All tornado queues get grouped into this, even if sharded
    "notify_tornado",
}

for queue_name in queues:
    queue_name = queue_name.strip()
    consumers[queue_name] = 0

for line in output.split("\n"):
    parts = line.split("\t")
    if len(parts) >= 2:
        queue_name = parts[0]
        if queue_name.startswith("notify_tornado_"):
            queue_name = "notify_tornado"
        consumers[queue_name] += 1

now = int(time.time())

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

    target_count = 1
    if queue_name == "notify_tornado":
        target_count = TORNADO_PROCESSES

    if consumers[queue_name] < target_count:
        status = 2
    else:
        status = 0
    with open(state_file_tmp, "w") as f:
        f.write(
            f"{now}|{status}|{states[status]}|queue {queue_name} has {consumers[queue_name]} consumers, needs {target_count}\n"
        )
    os.rename(state_file_tmp, state_file_path)