You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

142 lines
4.4 KiB

5 years ago
#!/usr/bin/python
# -*- coding: utf-8 -*-
# @Time : 2019-10-15
# @Author : lework
# @Desc : 收集supervisor的进程状态信息,并将信息暴露给Prometheus。
# [program:supervisor_exporter]
# process_name=%(program_name)s
# command=/usr/bin/python /root/scripts/supervisor_exporter.py
# autostart=true
# autorestart=true
# redirect_stderr=true
# stdout_logfile=/var/log/supervisor/supervisor_exporter.log
# stdout_logfile_maxbytes=50MB
# stdout_logfile_backups=3
# buffer_size=10
5 years ago
import sys
5 years ago
from time import sleep
5 years ago
from supervisor.xmlrpc import SupervisorTransport
from prometheus_client import Gauge, Counter, CollectorRegistry ,generate_latest, start_http_server
5 years ago
5 years ago
PY2 = sys.version_info[0] == 2
PY3 = sys.version_info[0] == 3
if PY3:
from xmlrpc.client import Transport, ServerProxy, Fault
5 years ago
from http.server import BaseHTTPRequestHandler, HTTPServer
5 years ago
else:
from xmlrpclib import Transport, ServerProxy, Fault
from BaseHTTPServer import BaseHTTPRequestHandler, HTTPServer
def get_supervisord_conn(supervisord_url, supervisord_user, supervisord_pass):
"""
获取supervisor的连接
:return:
"""
transport = SupervisorTransport(supervisord_user, supervisord_pass, supervisord_url)
s = ServerProxy('http://127.0.0.1', transport=transport)
return s
5 years ago
def is_runing(state):
state_info = {
5 years ago
# 'STOPPED': 0,
'STARTING': 10,
'RUNNING': 20
# 'BACKOFF': 30,
# 'STOPPING': 40
# 'EXITED': 100,
# 'FATAL': 200,
# 'UNKNOWN': 1000
5 years ago
}
if state in state_info.values():
return True
return False
def get_metrics():
collect_reg = CollectorRegistry(auto_describe=True)
try:
5 years ago
s = get_supervisord_conn(supervisord_url, supervisord_user, supervisord_pass)
5 years ago
data = s.supervisor.getAllProcessInfo()
except Exception as e:
print("unable to call supervisord: %s" % e)
return collect_reg
labels=('name', 'group')
metric_state = Gauge('state', "Process State", labelnames=labels, subsystem='supervisord', registry=collect_reg)
metric_exit_status=Gauge('exit_status', "Process Exit Status", labelnames=labels, subsystem='supervisord', registry=collect_reg)
metric_up = Gauge('up', "Process Up", labelnames=labels, subsystem='supervisord', registry=collect_reg)
metric_start_time_seconds=Counter('start_time_seconds', "Process start time", labelnames=labels, subsystem='supervisord', registry=collect_reg)
for item in data:
now = item.get('now', '')
group = item.get('group', '')
description = item.get('description', '')
stderr_logfile = item.get('stderr_logfile', '')
stop = item.get('stop', '')
statename = item.get('statename', '')
start = item.get('start', '')
state = item.get('state', '')
stdout_logfile = item.get('stdout_logfile', '')
logfile = item.get('logfile', '')
spawnerr = item.get('spawnerr', '')
name = item.get('name', '')
exitstatus = item.get('exitstatus', '')
labels = (name, group)
metric_state.labels(*labels).set(state)
metric_exit_status.labels(*labels).set(exitstatus)
if is_runing(state):
metric_up.labels(*labels).set(1)
5 years ago
metric_start_time_seconds.labels(*labels).inc(start)
5 years ago
else:
metric_up.labels(*labels).set(0)
5 years ago
return collect_reg
5 years ago
class myHandler(BaseHTTPRequestHandler):
def do_GET(self):
self.send_response(200)
self.send_header('Content-type','text/plain')
self.end_headers()
data=""
if self.path=="/":
5 years ago
data=b"hello, supervistor.\r\n\r\n/metrics"
elif self.path=="/metrics":
5 years ago
data=generate_latest(get_metrics())
else:
5 years ago
data=b"not found"
5 years ago
# Send the html message
self.wfile.write(data)
5 years ago
5 years ago
return
if __name__ == '__main__':
try:
5 years ago
supervisord_url = "unix:///var/run/supervisor.sock"
supervisord_user = ""
supervisord_pass = ""
5 years ago
5 years ago
PORT_NUMBER = 8081
5 years ago
#Create a web server and define the handler to manage the
#incoming request
server = HTTPServer(('', PORT_NUMBER), myHandler)
5 years ago
print('Started httpserver on port',PORT_NUMBER)
5 years ago
#Wait forever for incoming htto requests
server.serve_forever()
except KeyboardInterrupt:
5 years ago
print('^C received, shutting down the web server')
5 years ago
server.socket.close()