Kafak监控

推荐的监控工具

kafka-offset-monitor

Python实现监控

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
topics = ['heimdallr-dev']
# 要监控的groupid
monitor_group_ids = ['heimdallr']
# broker-list
servers = 'localhost:9092'
# 监控数据上报间隔 秒
time_interval = 10
# 历史全量数据上报间隔
history_time_interval = 5 * 60
# -*- coding:utf-8 -*-
import time

import sys

from kafka.client import KafkaClient
from kafka.protocol.commit import OffsetFetchRequest_v1, OffsetFetchResponse_v1, OffsetFetchRequest_v0, \
OffsetFetchResponse_v0
from kafka.protocol.offset import OffsetRequest_v0, OffsetResponse_v0

#from monitor_constants import *

duration = 0
client = None
conn = None
partition_cache = {}
brokers_cache = []
kafka_type = ['heimdallr']
zk_type = []


def get_brokers():
if not brokers_cache:
brokers = client.cluster.brokers()
if brokers:
brokers_cache.extend([x.nodeId for x in brokers])
return brokers_cache


def get_partitions(topic):
if not partition_cache or topic not in partition_cache:
partitions = client.cluster.available_partitions_for_topic(topic)
if partitions:
partition_cache[topic] = [x for x in partitions]
else:
return []
return partition_cache[topic]


def get_logsize():
"""
获取topic 下每个partition的logsize(各个broker的累加)
:return:
"""
tp = {} # topic : partition_dict
brokers = get_brokers()
for topic in topics:
partitions = get_partitions(topic)
pl = {} # partition : logsize
for broker in brokers:
# 这里取笛卡尔积可能有问题,但是不影响parse中解析了
for partition in partitions:
client.send(broker, OffsetRequest_v0(replica_id=-1, topics=[(topic, [(partition, -1, 1)])]))
responses = client.poll()
pdict = parse_logsize(topic, partition, responses)
append(pl, pdict)
tp[topic] = pl
return tp


def append(rdict, pdict):
if rdict:
# 已经有记录,累加
for k, v in pdict.items():
if k in rdict:
rdict[k] = rdict[k] + v
else:
rdict[k] = v
else:
rdict.update(pdict)


def parse_logsize(t, p, responses):
"""
单个broker中单个partition的logsize
:param responses:
:param p:
:param t:
:return:
"""
for response in responses:
if not isinstance(response, OffsetResponse_v0):
return {}
tps = response.topics
topic = tps[0][0]
partition_list = tps[0][1]
partition = partition_list[0][0]
# 异步poll来的数据可能不准
if topic == t and partition == p and partition_list[0][1] == 0:
logsize_list = partition_list[0][2]
logsize = logsize_list[0]
return {partition: logsize}
return {}


def parse_offsets(t, responses):
dr = {}
for response in responses:
if not isinstance(response, (OffsetFetchResponse_v1, OffsetFetchResponse_v0)):
return {}
tps = response.topics
topic = tps[0][0]
partition_list = tps[0][1]
if topic == t:
for partition_tunple in partition_list:
if partition_tunple[3] == 0:
offset = partition_tunple[1]
dr[partition_tunple[0]] = offset
return dr


def get_offsets():
# {gid: dict}
gd = {}
for gid in monitor_group_ids:
td = {} # {topic:dict}
for topic in topics:
pd = {} # {partition:dict}
for broker in get_brokers():
partitions = get_partitions(topic)
if not partitions:
return {}
else:
responses = optionnal_send(broker, gid, topic, partitions)
dr = parse_offsets(topic, responses)
append(pd, dr)
td[topic] = pd
gd[gid] = td
return gd


def optionnal_send(broker, gid, topic, partitions):
if gid in kafka_type:
return kafka_send(broker, gid, topic, partitions)
elif gid in zk_type:
return zk_send(broker, gid, topic, partitions)
else:
responses = zk_send(broker, gid, topic, partitions)
dct = parse_offsets(topic, responses)
if is_suitable(dct):
zk_type.append(gid)
return responses
responses = kafka_send(broker, gid, topic, partitions)
dct = parse_offsets(topic, responses)
if is_suitable(dct):
kafka_type.append(gid)
return responses


def is_suitable(dct):
for x in dct.values():
if x != -1:
return True


def kafka_send(broker, gid, topic, partitions):
client.send(broker, OffsetFetchRequest_v1(consumer_group=gid, topics=[(topic, partitions)]))
return client.poll()


def zk_send(broker, gid, topic, partitions):
client.send(broker, OffsetFetchRequest_v0(consumer_group=gid, topics=[(topic, partitions)]))
return client.poll()

def do_task():
offset_dict = get_offsets()
#print (offset_dict)
logsize_dict = get_logsize()
#print (logsize_dict)
print ('----------kafka monitor, info:-------------')
for gk, gv in offset_dict.items():
for tk, tv in gv.items():
for pk, pv in tv.items():
if logsize_dict and tk in logsize_dict:
dr = logsize_dict[tk] # partition:logsize
if dr and pk in dr:
param = (gk, tk, pk, pv, dr[pk],
time.strftime('%Y-%m-%d %H:%M:%S', time.localtime(time.time())))
print(param)


if __name__ == "__main__":
client = KafkaClient(bootstrap_servers=servers, request_timeout_ms=3000)
while True:
do_task()
time.sleep(time_interval)
duration += time_interval

参考:https://my.oschina.net/ktlb/blog/863308