目录
- 监控需求
- 监控脚本
- 监控界面
- 总结
监控需求
公司部署了套redis集群，三主三从，虽然是高可用，但是还是要做好监控，即时发现问题。网上虽然有zabbix的redis监控模板，监控项不够简练，于是自己用python写了个。
监控脚本
pip install rediscluster
#!/usr/local/python/shims/python
from rediscluster import StrictRedisCluster
import redis
import sys
import mylog
redis_nodes = [{'host':'ip','port':port},
{'host':'ip','port':port},
{'host':'ip','port':port},
{'host':'ip','port':port},
{'host':'ip','port':port},
{'host':'ip','port':port},
]
###
进入redis集群模式，如果异常，记录到日志中，并终止脚本
###
try:
redisconn = StrictRedisCluster(startup_nodes=redis_nodes,password='pwd')
except Exception,e:
mylog.logging.error('%s' %e )
sys.exit(0)
###
定义参数
###
data = {}
NodeData = {}
hit = 0
misshit = 0
hitrate = 0.00
###
定义函数，抓取监控项
###
def ClusterState(item):
cluster_state = redisconn.execute_command('cluster','info')
cluster_state = cluster_state.split('\r\n')
try:
for i in cluster_state:
data[i.split(':')[0]] = i.split(':')[1]
except:
pass
if item == 'clusterstatus':
state = data['cluster_state']
if state == 'ok' :
item = 1
else:
item = 0
return item
elif item == 'clusterslotsfail':
item = data['cluster_slots_fail']
return item
elif item == 'clusterknownnodes':
item = data['cluster_known_nodes']
return item
else:
return 9999
def NodeInfoServer(item):
node_info = redisconn.info('Server')
NodeData = node_info['ip:port']
if item == 'uptime_in_days':
item = NodeData['uptime_in_days']
return item
else:
return 9999
def NodeInfoClients(item):
node_info = redisconn.info('Clients')
NodeData = node_info['ip:port']
if item == 'connected_clients':
item = NodeData['connected_clients']
return item
else:
return 9999
def NodeInfoMemory(item):
node_info = redisconn.info('Memory')
NodeData = node_info['ip:port']
if item == 'used_memory_human':
item = NodeData['used_memory_human']
return item
elif item == 'total_system_memory_human':
item = NodeData['total_system_memory_human']
return item
else:
return 9999
def NodeInfoPersistence(item):
node_info = redisconn.info('Persistence')
NodeData = node_info['ip:port']
if item == 'rdb_last_bgsave_status':
item = NodeData['rdb_last_bgsave_status']
if item == 'ok' :
item = 1
else:
item = 0
return item
else:
return 9999
def NodeInfoStats(item):
node_info = redisconn.info('Stats')
NodeData = node_info['ip:port']
if item == 'instantaneous_ops_per_sec':
item = NodeData['instantaneous_ops_per_sec']
return item
elif item == 'instantaneous_input_kbps':
item = NodeData['instantaneous_input_kbps']
return item
elif item == 'instantaneous_output_kbps':
item = NodeData['instantaneous_output_kbps']
return item
elif item == 'hit':
hit = NodeData['keyspace_hits']
misshit = NodeData['keyspace_misses']
hitrate = round((float(hit) / float(hit + misshit)) ,3)
item = hitrate
return item
else:
return 9999
###
脚本传参，zabbix获取监控项
###
if sys.argv[1] == 'status':
print ClusterState('clusterstatus')
elif sys.argv[1] == 'slotsfail':
print ClusterState('clusterslotsfail')
elif sys.argv[1] == 'nodes':
print ClusterState('clusterknownnodes')
elif sys.argv[1] == 'day':
print NodeInfoServer('uptime_in_days')
elif sys.argv[1] == 'clients':
print NodeInfoClients('connected_clients')
elif sys.argv[1] == 'usememory':
print NodeInfoMemory('used_memory_human')
elif sys.argv[1] == 'sysmemory':
print NodeInfoMemory('total_system_memory_human')
elif sys.argv[1] == 'rdb':
print NodeInfoPersistence('rdb_last_bgsave_status')
elif sys.argv[1] == 'ops':
print NodeInfoStats('instantaneous_ops_per_sec')
elif sys.argv[1] == 'input_kbps':
print NodeInfoStats('instantaneous_input_kbps')
elif sys.argv[1] == 'output_kbps':
print NodeInfoStats('instantaneous_output_kbps')
elif sys.argv[1] == 'hit':
print NodeInfoStats('hit')
监控界面
1513862328(1).jpg
监控的参数有input_kps、output_kps、ops、redis运行天数、占用系统内存、失败的槽数、客户端数、持久化状态、缓存命中率、集群状态等等，结合生产情况设置阈值告警
总结
由于个人对redis理解有限，暂时只监控了这些参数。python写的也不够简洁，很多需要优化的地方。先暂时实现了功能，有空再优化