一般用官方的rabbitmq_exporter采取数据即可,然后在普米配置。但如果rabbitmq节点的队列数超过了5000,往往rabbitmq_exporter就会瘫痪,因为rabbitmq_exporter采集的信息太多,尤其是那些队列的细节,所以队列多了,rabbitmq_exporter就没法用了。所以我们不得不自己写脚本探测MQ,脚本分享如下:
首先 pip3 install prometheus-client
import prometheus_client as prom
import pandas as pd
from sqlalchemy import create_engine
import requests,time
#自定义普米MQ监控指标
port = '15672'
username = 'username'
password = 'password'
g0 = prom.Gauge("rabbitmq_up", 'life of the node',labelnames=['node','region'])
g1 = prom.Gauge("rabbitmq_queues", 'total queue num of the node',labelnames=['node','region'])
g2 = prom.Gauge("rabbitmq_channels", 'total queue num of the node',labelnames=['node','region'])
g3 = prom.Gauge("rabbitmq_connections", 'total queue num of the node',labelnames=['node','region'])
g4 = prom.Gauge("rabbitmq_consumers", 'total queue num of the node',labelnames=['node','region'])
g5 = prom.Gauge("rabbitmq_exchanges", 'total queue num of the node',labelnames=['node','region'])
g6 = prom.Gauge("rabbitmq_messages", 'total messages of the node',labelnames=['node','region'])
g7 = prom.Gauge("rabbitmq_vhosts", 'total vhost num of the node',labelnames=['node','region'])
g8 = prom.Gauge("rabbitmq_node_mem_used", 'mem used of the node',labelnames=['node','region'])
g9 = prom.Gauge("rabbitmq_node_mem_limit", 'mem limit of the node',labelnames=['node','region'])
g10 = prom.Gauge("rabbitmq_node_mem_alarm", 'mem alarm of the node',labelnames=['node','region'])
g11 = prom.Gauge("rabbitmq_node_disk_free_alarm", 'free disk alarm of the node',labelnames=['node','region'])
prom.start_http_server(8086)
#要监控的MQ节点
nodelist=['1.1.1.1','1.1.1.2','1.1.1.3']
while True:
for node in nodelist: #遍历各个node
status=1
try: #测试连通性
requests.get(url=f"http://{node}:{port}/api/overview", auth=(username, password),timeout=5)
except:
status=0
continue
finally:
g0.labels(node=node,region=region).set(status)
info1=requests.get(url=f"http://{node}:{port}/api/overview", auth=(username, password),timeout=5)
info2=requests.get(url=f"http://{node}:{port}/api/nodes", auth=(username, password),timeout=5)[0]
info3=requests.get(url=f"http://{node}:{port}/api/vhosts", auth=(username, password),timeout=5)
g1.labels(node=node,region=region).set(info1.get('object_totals').get('queues'))
g2.labels(node=node,region=region).set(info1.get('object_totals').get('channels'))
g3.labels(node=node,region=region).set(info1.get('object_totals').get('connections'))
g4.labels(node=node,region=region).set(info1.get('object_totals').get('consumers'))
g5.labels(node=node,region=region).set(info1.get('object_totals').get('exchanges'))
g6.labels(node=node,region=region).set(info1.get('queue_totals').get('messages'))
g7.labels(node=node,region=region).set(len(info3))
g8.labels(node=node,region=region).set(info2.get('mem_used'))
g9.labels(node=node,region=region).set(info2.get('mem_limit'))
g10.labels(node=node,region=region).set(info2.get('mem_alarm'))
g11.labels(node=node,region=region).set(info2.get('disk_free_alarm'))
time.sleep(30)
python3 执行这个脚本,就会运行一个页面如下
于是就可以用普米采集了
注意honor_labels 指标,就是让job,instance指标不被普米自身的覆盖
或者写脚本推送到Pushgateway也可以,如下
import prometheus_client as prom
import pandas as pd
from sqlalchemy import create_engine
import requests,time
#自定义普米MQ监控指标
port = '15672'
username = 'username'
password = 'password'
url='http://pushgateway地址:9091'
registry = prom.CollectorRegistry()
job='rabbitmq'
g0 = prom.Gauge("rabbitmq_up", 'life of the node',labelnames=['node','region'],registry=registry)
g1 = prom.Gauge("rabbitmq_queues", 'total queue num of the node',labelnames=['node','region'],registry=registry)
g2 = prom.Gauge("rabbitmq_channels", 'total queue num of the node',labelnames=['node','region'],registry=registry)
g3 = prom.Gauge("rabbitmq_connections", 'total queue num of the node',labelnames=['node','region'],registry=registry)
g4 = prom.Gauge("rabbitmq_consumers", 'total queue num of the node',labelnames=['node','region'],registry=registry)
g5 = prom.Gauge("rabbitmq_exchanges", 'total queue num of the node',labelnames=['node','region'],registry=registry)
g6 = prom.Gauge("rabbitmq_messages", 'total messages of the node',labelnames=['node','region'],registry=registry)
g7 = prom.Gauge("rabbitmq_vhosts", 'total vhost num of the node',labelnames=['node','region'],registry=registry)
g8 = prom.Gauge("rabbitmq_node_mem_used", 'mem used of the node',labelnames=['node','region'],registry=registry)
g9 = prom.Gauge("rabbitmq_node_mem_limit", 'mem limit of the node',labelnames=['node','region'],registry=registry)
g10 = prom.Gauge("rabbitmq_node_mem_alarm", 'mem alarm of the node',labelnames=['node','region'],registry=registry)
g11 = prom.Gauge("rabbitmq_node_disk_free_alarm", 'free disk alarm of the node',labelnames=['node','region'],registry=registry)
#要监控的MQ节点
nodelist=['1.1.1.1','1.1.1.2','1.1.1.3']
while True:
for node in nodelist: #遍历各个node
status=1
try: #测试连通性
requests.get(url=f"http://{node}:{port}/api/overview", auth=(username, password),timeout=5)
except:
status=0
continue
finally:
g0.labels(node=node,region=region).set(status)
info1=requests.get(url=f"http://{node}:{port}/api/overview", auth=(username, password),timeout=5)
info2=requests.get(url=f"http://{node}:{port}/api/nodes", auth=(username, password),timeout=5)[0]
info3=requests.get(url=f"http://{node}:{port}/api/vhosts", auth=(username, password),timeout=5)
g1.labels(node=node,region=region).set(info1.get('object_totals').get('queues'))
g2.labels(node=node,region=region).set(info1.get('object_totals').get('channels'))
g3.labels(node=node,region=region).set(info1.get('object_totals').get('connections'))
g4.labels(node=node,region=region).set(info1.get('object_totals').get('consumers'))
g5.labels(node=node,region=region).set(info1.get('object_totals').get('exchanges'))
g6.labels(node=node,region=region).set(info1.get('queue_totals').get('messages'))
g7.labels(node=node,region=region).set(len(info3))
g8.labels(node=node,region=region).set(info2.get('mem_used'))
g9.labels(node=node,region=region).set(info2.get('mem_limit'))
g10.labels(node=node,region=region).set(info2.get('mem_alarm'))
g11.labels(node=node,region=region).set(info2.get('disk_free_alarm'))
prom.push_to_gateway(url,job=job,registry=registry, timeout=20)
time.sleep(30)