JMX监控kafka各种指标
nohup KAFKA_JMX_OPTS="-Dcom.sun.management.jmxremote=true -Dcom.sun.management.jmxremote.authenticate=fal -Dcom.sun.management.jmxremote.ssl=fal -i.rver.hostname=192.168.71.25 -Djava.preferIPv4Stack=true -Dcom.sun.management.jmxremote.port=9999" ./kafka-rver-start.sh …/config/rver.properties 1>/dev/null 2>&1 &
Kafka常⽤JMX监控指标整理【实战笔记】
⽬录
⼀、系统相关指标
⼆、GC相关指标
三、JVM相关指标
四、Topic相关指标
五、Broker相关指标
六、系列⽂章
⼀、系统相关指标
1.系统信息收集
java.lang:type=OperatingSystem
看见读后感{“freePhysicalMemorySize”:“806023168”,“maxFileDescriptorCount”:“4096”,“openFileDescriptorCount”:“283”,“processCpuLoad”:“0.0017562901839817224”,“systemCpuLoad”:“0.014336627412954635”,“systemLoadAv erage”:“0.37”}
2.Thread信息收集
java.lang:type=Threading
{“peakThreadCount”:“88”,“threadCount”:“74”}
3.获取mmaped和direct空间
通过BufferPoolMXBean获取ud、capacity、count
⼆、GC相关指标
1.Young GC
java.lang:type=GarbageCollector,name=G1 Young Generation
{“collectionCount”:“534”,“collectionTime”:“8258”}
2.Old GC
java.lang:type=GarbageCollector,name=G1 Old Generation
{“collectionCount”:“0”,“collectionTime”:“0”}
三、JVM相关指标
通过MemoryMXBean获取JVM相关信息HeapMemoryUsage和NonHeapMemoryUsage；通过MemoryPoolMXBean获取其他JVM 内存空间指标，例如：Metaspace、Codespace等
四、Topic相关指标中秋节晚会开场白
1.Topic消息⼊站速率（Byte）
kafka.rver:type=BrokerTopicMetrics,name=BytesInPerSec,topic=" + topic
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
2.Topic消息出站速率（Byte）
kafka.rver:type=BrokerTopicMetrics,name=BytesOutPerSec,topic=" + topic
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
3.Topic请求被拒速率
kafka.rver:type=BrokerTopicMetrics,name=BytesRejectedPerSec,topic=" + topic
4.Topic失败拉去请求速率
kafka.rver:type=BrokerTopicMetrics,name=FailedFetchRequestsPerSec,topic=" + topic;
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
5.Topic发送请求失败速率
kafka.rver:type=BrokerTopicMetrics,name=FailedProduceRequestsPerSec,topic=" + topic
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
6.Topic消息⼊站速率（message）
kafka.rver:type=BrokerTopicMetrics,name=MessagesInPerSec,topic=" + topic
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
五、Broker相关指标
1.Log flush rate and time
kafka.log:type=LogFlushStats,name=LogFlushRateAndTimeMs
{“50thPercentile”:“1.074103”,“75thPercentile”:“1.669793”,“95thPercentile”:“6.846556”,“98thPercentile”:“6.846556”,“999thPercentile”:“6.846556”,“99thPercentile”:“6.846556”,“count”:“19”,“max”:“6.846556”,“mean”:“1.628646052631579”,“min”:“0.512879”,“stdDev”:“1.6007003364105892”}
2.同步失效的副本数
kafka.rver:type=ReplicaManager,name=UnderReplicatedPartitions
{“value”:“0”}
3.消息⼊站速率（消息数）
kafka.rver:type=BrokerTopicMetrics,name=MessagesInPerSec
{“count”:“86845”,“fifteenMinuteRate”:“0.6456600497006455”,“fiveMinuteRate”:“0.6444164288097876”,“meanRate”:“0.5314899330400695”,“oneMinuteRate”:“0.6494649408329609”}
4.消息⼊站速率（Byte）
kafka.rver:type=BrokerTopicMetrics,name=BytesInPerSec
{“count”:“57302357”,“fifteenMinuteRate”:“379.11342092748146”,“fiveMinuteRate”:“371.8482236385939”,“meanRate”:“351.37122686037435”,“oneMinuteRate”:“351.8348952308101”}
5.消息出站速率（Byte）
kafka.rver:type=BrokerTopicMetrics,name=BytesOutPerSec
{“count”:“246”,“fifteenMinuteRate”:“4.508738367219028E-34”,“fiveMinuteRate”:“1.4721921790135324E-98”,“meanRate”:“0.0015031168286836175”,“oneMinuteRate”:“2.964393875E-314”}
6.请求被拒速率
kafka.rver:type=BrokerTopicMetrics,name=BytesRejectedPerSec
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
7.失败拉去请求速率
kafka.rver:type=BrokerTopicMetrics,name=FailedFetchRequestsPerSec
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
8.发送请求失败速率
kafka.rver:type=BrokerTopicMetrics,name=FailedProduceRequestsPerSec
9.Leader副本数
kafka.rver:type=ReplicaManager,name=LeaderCount菜名字大全
{“value”:“92”}
10.Partition数量
kafka.rver:type=ReplicaManager,name=PartitionCount
{“value”:“135”}
11.下线Partition数量
{“value”:“0”}
12.Broker⽹络处理线程空闲率
kafka.rver:type=KafkaRequestHandlerPool,name=RequestHandlerAvgIdlePercent
达濠鱼丸
{“count”:“164506926671008”,“fifteenMinuteRate”:“0.9999327359820058”,“fiveMinuteRate”:“1.000029005 4537715”,“meanRate”:“0.9998854371393514”,“oneMinuteRate”:“1.0007836499581673”}
13.Leader选举⽐率
{“count”:“7”,“fifteenMinuteRate”:“5.134993718576819E-82”,“fiveMinuteRate”:“6.882658450509451E-240”,“meanRate”:“4.2525243043608314E-5”,“oneMinuteRate”:“2.964393875E-314”}
14.Unclean Leader选举⽐率
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:“0.0”}
15.Controller存活数量
新年贺词集锦
{“value”:“1”}
16.请求速率
kafkawork:type=RequestMetrics,name=RequestsPerSec,request=Produce
{“count”:“83233”,“fifteenMinuteRate”:“0.6303485369828705”,“fiveMinuteRate”:“0.6357199085092445”,“meanRate”:“0.5046486472186744”,“oneMinuteRate”:“0.6563203475530601”}
17.Consumer拉取速率
kafkawork:type=RequestMetrics,name=RequestsPerSec,request=FetchConsumer
{“count”:“125796”,“fifteenMinuteRate”:“1.14193044007404E-33”,“fiveMinuteRate”:“7.699516480260211E-100”,“meanRate”:“0.7623419964866819”,“oneMinuteRate”:“2.964393875E-314”}
青角18.Follower拉去速率
kafkawork:type=RequestMetrics,name=RequestsPerSec,request=FetchFollower
{“count”:“375108”,“fifteenMinuteRate”:“2.302746562040189”,“fiveMinuteRate”:“2.292459728166488”,“meanRate”:“2.2721808581484693”,“oneMinuteRate”:“2.2814260196672973”}
19.Request total time
kafkawork:type=RequestMetrics,name=TotalTimeMs,request=Produce
{“50thPercentile”:“1.0”,“75thPercentile”:“1.0”,“95thPercentile”:“2.0”,“98thPercentile”:“2.0”,“999thPerce ntile”:“28.0”,“99thPercentile”:“4.0”,“count”:“83384”,“max”:“48.0”,“mean”:“1.2344934279957787”,“min”:“0.0”,“stdDev”:“1.1783192073287214”}
20.Consumer fetch total time
kafkawork:type=RequestMetrics,name=TotalTimeMs,request=FetchConsumer
{“50thPercentile”:“500.0”,“75thPercentile”:“501.0”,“95thPercentile”:“501.0”,“98thPercentile”:“501.0”,“9 99thPercentile”:“501.971”,“99thPercentile”:“501.0”,“count”:“125796”,“max”:“535.0”,“mean”:“499.831 23469744663”,“min”:“0.0”,“stdDev”:“17.138716708632025”}
21.Follower fetch total time
kafkawork:type=RequestMetrics,name=TotalTimeMs,request=FetchFollower
{“50thPercentile”:“500.0”,“75thPercentile”:“500.0”,“95thPercentile”:“501.0”,“98thPercentile”:“501.0”,“9 99thPercentile”:“507.826”,“99thPercentile”:“501.0”,“count”:“375564”,“max”:“532.0”,“mean”:“437.797 63502359117”,“min”:“0.0”,“stdDev”:“148.25999023472986”}我的自白
22.Time the follower fetch request waits in the request queue
kafkawork:type=RequestMetrics,name=RequestQueueTimeMs,request=FetchFollower
{“50thPercentile”:“0.0”,“75thPercentile”:“0.0”,“95thPercentile”:“0.0”,“98thPercentile”:“0.0”,“999thPerce ntile”:“0.0”,“99thPercentile”:“0.0”,“count”:“376206”,“max”:“28.0”,“mean”:“0.0010260336092459982”,“min”:“0.0”,“stdDev”:“0.1282889653905258”}
23.Time the Consumer fetch request waits in the request queue
kafkawork:type=RequestMetrics,name=RequestQueueTimeMs,request=FetchConsumer
{“50thPercentile”:“0.0”,“75thPercentile”:“0.0”,“95thPercentile”:“0.0”,“98thPercentile”:“0.0”,“999thPerce ntile”:“0.0”,“99thPercentile”:“0.0”,“count”:“125796”,“max”:“24.0”,“mean”:“0.0018124582657636174”,“min”:“0.0”,“stdDev”:“0.18122860552537737”}
24.Time the Produce fetch request waits in the request queue
kafkawork:type=RequestMetrics,name=RequestQueueTimeMs,request=Produce
{“50thPercentile”:“0.0”,“75thPercentile”:“0.0”,“95thPercentile”:“0.0”,“98thPercentile”:“0.0”,“999thPerce ntile”:“0.0”,“99thPercentile”:“0.0”,“count”:“83704”,“max”:“12.0”,“mean”:“2.6283092803211315E-4”,“min”:“0.0”,“stdDev”:“0.042892540270754634”}
25.Broker I/O⼯作处理线程空闲率
kafkawork:type=SocketServer,name=NetworkProcessorAvgIdlePercent
{“value”:“1.0015540075894207”}
26.ISR变化速率
kafka.rver:type=ReplicaManager,name=IsrShrinksPerSec
{“count”:“0”,“fifteenMinuteRate”:“0.0”,“fiveMinuteRate”:“0.0”,“meanRate”:“0.0”,“oneMinuteRate”:"0
开启JMX端⼝
修改bin/kafka-rver-start.sh，添加JMX_PORT参数，添加后样⼦如下
if [ “x$KAFKA_HEAP_OPTS” = “x” ]; then
export KAFKA_HEAP_OPTS="-Xmx1G -Xms1G"
export JMX_PORT=“9999”
fi
通过Jconsole测试时候可以连接
通过JavaAPI来访问
通过以下⽅法获取⽬标值
public class KafkaDataProvider{
protected final Logger LOGGER = Logger(getClass());
private static final String MESSAGE_IN_PER_SEC = “kafka.rver:type=BrokerTopicMetrics,name=MessagesInPerSec”; private static final String BYTES_IN_PER_SEC = “kafka.rver:type=BrokerTopicMetrics,name=BytesInPerSec”;
private static final String BYTES_OUT_PER_SEC = “kafka.rver:type=BrokerTopicMetrics,name=BytesOutPerSec”;
private static final String PRODUCE_REQUEST_PER_SEC =
“kafkawork:type=RequestMetrics,name=RequestsPerSec,request=Produce”;
private static final String CONSUMER_REQUEST_PER_SEC =
“kafkawork:type=RequestMetrics,name=RequestsPerSec,request=FetchConsumer”;
private static final String FLOWER_REQUEST_PER_SEC =
“kafkawork:type=RequestMetrics,name=RequestsPerSec,request=FetchFollower”;
private static final String ACTIVE_CONTROLLER_COUNT =
鼻翼长痘
“ller:type=KafkaController,name=ActiveControllerCount”;
private static final String PART_COUNT = “kafka.rver:type=ReplicaManager,name=PartitionCount”;
public String extractMonitorData() {
//TODO 通过调⽤API获得IP以及参数
KafkaRoleInfo monitorDataPoint = new KafkaRoleInfo();
String jmxURL = “rvice:jmx:rmi:///jndi/rmi://192.168.40.242:9999/jmxrmi”;
try {
MBeanServerConnection jmxConnection = MBeanServerConnection(jmxURL);
ObjectName messageCountObj = new ObjectName(MESSAGE_IN_PER_SEC);
ObjectName bytesInPerSecObj = new ObjectName(BYTES_IN_PER_SEC);
ObjectName bytesOutPerSecObj = new ObjectName(BYTES_OUT_PER_SEC);
ObjectName produceRequestsPerSecObj = new ObjectName(PRODUCE_REQUEST_PER_SEC);
ObjectName consumerRequestsPerSecObj = new ObjectName(CONSUMER_REQUEST_PER_SEC);
ObjectName flowerRequestsPerSecObj = new ObjectName(FLOWER_REQUEST_PER_SEC);
ObjectName activeControllerCountObj = new ObjectName(ACTIVE_CONTROLLER_COUNT);
ObjectName partCountObj = new ObjectName(PART_COUNT);
Long messagesInPerSec = (Long) Attribute(messageCountObj, “Count”);
Long bytesInPerSec = (Long) Attribute(bytesInPerSecObj, “Count”);
Long bytesOutPerSec = (Long) Attribute(bytesOutPerSecObj, “Count”);
Long produceRequestCountPerSec = (Long) Attribute(produceRequestsPerSecObj, “Count”); Long consumerRequestCountPerSec = (Long) Attribute(consumerRequestsPerSecObj, “Count”); Long flowerRequestCountPerSec = (Long) Attribute(flowerRequestsPerSecObj, “Count”);
Integer activeControllerCount = (Integer) Attribute(activeControllerCountObj, “Value”);
Integer partCount = (Integer) Attribute(partCountObj, “Value”);
monitorDataPoint.tMessagesInPerSec(messagesInPerSec);
monitorDataPoint.tBytesInPerSec(bytesInPerSec);
monitorDataPoint.tBytesOutPerSec(bytesOutPerSec);
monitorDataPoint.tProduceRequestCountPerSec(produceRequestCountPerSec);
monitorDataPoint.tConsumerRequestCountPerSec(consumerRequestCountPerSec);
monitorDataPoint.tFlowerRequestCountPerSec(flowerRequestCountPerSec);
monitorDataPoint.tActiveControllerCount(activeControllerCount);
monitorDataPoint.tPartCount(partCount);
} catch (IOException e) {
e.printStackTrace();
} catch (MalformedObjectNameException e) {
e.printStackTrace();
} catch (AttributeNotFoundException e) {
e.printStackTrace();
} catch (MBeanException e) {
e.printStackTrace();
} catch (ReflectionException e) {
e.printStackTrace();
} catch (InstanceNotFoundException e) {
e.printStackTrace();
}
String();
}
public static void main(String[] args) {
System.out.println(new KafkaDataProvider().extractMonitorData());
}
/**
* 获得MBeanServer 的连接
*
* @param jmxUrl
