Prometheus指标
主要参数解释
# 节点IP和端口(instance)
例如:192.168.1.226:9100、192.168.1.227:9100、192.168.1.228:9100
# HDFS-DataNode-IP和端口
例如:192.168.1.226:30003、192.168.1.227:30003、192.168.1.228:30003
# Yarn-NodeManager-IP和端口
例如:192.168.1.226:30005、192.168.1.227:30005、192.168.1.228:30005
一、可用性监测(0代表存在异常或未启动,1代表运行中)
# 各个节点的可用性监测
up{job="node_exporter"}
# hdfs-namenode可用性监测
up{job="hdfs-namenode"}
# hdfs-datanode可用性监测
up{job="hdfs-datanode"}
# yarn-resourcemanager可用性监测
up{job="yarn-resourcemanager"}
# yarn-nodemanager可用性监测
up{job="yarn-nodemanager"}
# zookeeper可用性监测
up{job="zookeeper"}
# kafka可用性监测
up{job="kafka"}
# hiveserver2可用性监测
up{job="hiveserver2"}
# metastore可用性监测
up{job="metastore"}
- 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
二、节点监测
# 磁盘总容量
node_filesystem_size_bytes{instance="节点IP和端口",mountpoint="/"}/1024/1024/1024
# 磁盘可用容量
node_filesystem_avail_bytes{instance="节点IP和端口",mountpoint="/"} /1024 /1024/1024
# CPU负载
node_load1{instance="节点IP和端口"}
# CPU使用率
100-avg(irate(node_cpu_seconds_total{mode="idle",instance="节点IP和端口"}[1m])) by (instance)*100
# 内存饱和度
sum((rate(node_vmstat_pgpgin{instance="节点IP和端口"}[1m])+rate(node_vmstat_pgpgout{instance="节点IP和端口"}[1m]))) by (instance)
# 物理内存使用率
(node_memory_MemTotal_bytes{instance="节点IP和端口"} - (node_memory_MemFree_bytes{instance="节点IP和端口"} + node_memory_Buffers_bytes{instance="节点IP和端口"} + node_memory_Cached_bytes{instance="节点IP和端口"}))/node_memory_MemTotal_bytes{instance="节点IP和端口"} * 100
# SWAP内存使用率
(node_memory_SwapTotal_bytes{instance="节点IP和端口"} - node_memory_SwapFree_bytes{instance="节点IP和端口"})/node_memory_SwapTotal_bytes{instance="节点IP和端口"} * 100
# 磁盘分区使用率("/"分区)
(node_filesystem_size_bytes{mountpoint="/",instance="节点IP和端口"} - node_filesystem_free_bytes{mountpoint="/",instance="节点IP和端口"})/node_filesystem_size_bytes{mountpoint="/",instance="节点IP和端口"} * 100
# 磁盘吞吐量(每秒读速率)
irate(node_disk_read_bytes_total{device="sda",instance="节点IP和端口"}[1m]) / 1024
# 磁盘吞吐量(每秒写速率)
irate(node_disk_written_bytes_total{device="sda",instance="节点IP和端口"}[1m]) / 1024
# 磁盘IOPS(每秒读次数)
irate(node_disk_reads_completed_total{device="sda",instance="节点IP和端口"}[1m])
# 磁盘IOPS(每秒写次数)
irate(node_disk_writes_completed_total{device="sda",instance="节点IP和端口"}[1m])
# 网卡流量(下载流量总字节数)
irate(node_network_receive_bytes_total{device != "lo",device !~ "docker.*",instance="节点IP和端口"}[1m]) / 1024
# 网卡流量(上传流量总字节数)
irate(node_network_transmit_bytes_total{device != "lo",device !~ "docker.*",instance="节点IP和端口"}[1m]) / 1024
- 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
三、服务监测
1.HDFS监测
# HDFS NameNode JVM堆内存使用率
Hadoop_NameNode_MemHeapUsedM/Hadoop_NameNode_MemHeapMaxM * 100
# HDFS DataNode JVM堆内存使用率
Hadoop_DataNode_MemHeapUsedM{instance="HDFS-DataNode-IP和端口"}/Hadoop_DataNode_MemHeapMaxM{instance="HDFS-DataNode-IP和端口"} * 100
# HDFS 容量使用率(百分比)
Hadoop_NameNode_PercentUsed
# HDFS:CPU 使用率
sum(java_lang_OperatingSystem_ProcessCpuLoad{job=~"hdfs.*"})*100
# HDFS:NameNode RPC请求平均处理时间 (单位:ms)
Hadoop_NameNode_RpcProcessingTimeAvgTime{job="hdfs-namenode"}
# HDFS:NameNode RPC处理队列长度 (无单位或1/个)
Hadoop_NameNode_CallQueueLength{job="hdfs-namenode"}
# HDFS:DataNode RPC请求平均处理时间 (单位:ms)
Hadoop_DataNode_RpcProcessingTimeAvgTime{job="hdfs-datanode"}
# HDFS:DataNode RPC处理队列长度 (无单位或1/个)
Hadoop_DataNode_CallQueueLength{job="hdfs-datanode"}
# HDFS:NameNode 线程阻塞数量(无单位或1/个)
sum(Hadoop_NameNode_ThreadsBlocked{job="hdfs-namenode"})
# HDFS:NameNode 线程等待数量(无单位或1/个)
sum(Hadoop_NameNode_ThreadsWaiting{job="hdfs-namenode"})
# HDFS:DataNode 线程阻塞数量(无单位或1/个)
sum(Hadoop_DataNode_ThreadsBlocked{job="hdfs-datanode"})
# HDFS:DataNode 线程等待数量(无单位或1/个)
sum(Hadoop_DataNode_ThreadsWaiting{job="hdfs-datanode"})
# HDFS:DataNode 当前已使用容量(单位:MB)
Hadoop_NameNode_CapacityUsed{name="FSNamesystem"}/1024/1024
# HDFS:DataNode 原始总容量(单位:GB)
Hadoop_NameNode_CapacityTotal{name="FSNamesystem"}/1024/1024/1024
# HDFS:当前运行正常的DataNode数量
Hadoop_NameNode_NumLiveDataNodes
# HDFS:当前运行异常的DataNode数量
Hadoop_NameNode_NumDeadDataNodes
# HDFS:当前分配的block数量
Hadoop_NameNode_BlocksTotal{name="FSNamesystem"}
# HDFS:NameNode接收字节速率(单位:B/s)
irate(Hadoop_NameNode_ReceivedBytes[1m])
# HDFS:NameNode发送字节速率(单位:B/s)
irate(Hadoop_NameNode_SentBytes[1m])
# HDFS:NameNode创建文件的操作总数
Hadoop_NameNode_CreateFileOps
# HDFS:NameNode删除文件的操作总数
Hadoop_NameNode_DeleteFileOps
# HDFS:NameNode当前连接数
Hadoop_NameNode_TotalLoad{name="FSNamesystem"}
# HDFS:NameNode当前文件和目录的数量
Hadoop_NameNode_FilesTotal{name="FSNamesystem"}
# HDFS:DataNode读取字节速率(单位:B/s)
sum(irate(Hadoop_DataNode_BytesRead[1m]))
# HDFS:DataNode写入字节速率(单位:B/s)
sum(irate(Hadoop_DataNode_BytesWritten[1m]))
# HDFS:DataNode读操作总数
sum(Hadoop_DataNode_ReadBlockOpNumOps)
# HDFS:DataNode写操作总数
sum(Hadoop_DataNode_WriteBlockOpNumOps)
# HDFS:DataNode接收字节速率(单位:B/s)
sum(irate(Hadoop_DataNode_ReceivedBytes[1m]))
# HDFS:DataNode发送字节速率(单位:B/s)
sum(irate(Hadoop_DataNode_SentBytes[1m]))
- 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
2.Yarn监测
# Yarn ResourceManager JVM堆内存使用率
Hadoop_ResourceManager_MemHeapUsedM/Hadoop_ResourceManager_MemHeapMaxM * 100
# Yarn NodeManager JVM堆内存使用率
Hadoop_NodeManager_MemHeapUsedM{instance="Yarn-NodeManager-IP和端口"}/Hadoop_NodeManager_MemHeapMaxM{instance="Yarn-NodeManager-IP和端口"} * 100
# Yarn:CPU 使用率
sum(java_lang_OperatingSystem_ProcessCpuLoad{job=~"yarn.*"})*100
# Yarn:ResourceManager RPC请求平均处理时间(单位:ms)
Hadoop_ResourceManager_RpcProcessingTimeAvgTime{job="yarn-resourcemanager"}
# Yarn:ResourceManager RPC处理队列长度(无单位或1/个)
Hadoop_ResourceManager_CallQueueLength{job="yarn-resourcemanager"}
# Yarn:NodeManager RPC请求平均处理时间(单位:ms)
Hadoop_NodeManager_RpcProcessingTimeAvgTime{job="yarn-nodemanager"}
# Yarn:NodeManager RPC处理队列长度(无单位或1/个)
Hadoop_NodeManager_CallQueueLength{job="yarn-nodemanager"}
# Yarn:ResourceManager 线程阻塞数量(无单位或1/个)
sum(Hadoop_ResourceManager_ThreadsBlocked{job="yarn-resourcemanager"})
# Yarn:NodeManager 线程阻塞数量(无单位或1/个)
sum(Hadoop_NodeManager_ThreadsBlocked{job="yarn-nodemanager"})
# Yarn:ResourceManager 线程等待数量(无单位或1/个)
sum(Hadoop_ResourceManager_ThreadsWaiting{job="yarn-resourcemanager"})
# Yarn:NodeManager 线程等待数量(无单位或1/个)
sum(Hadoop_NodeManager_ThreadsWaiting{job="yarn-nodemanager"})
# Yarn:NM存活节点数量统计
Hadoop_ResourceManager_NumActiveNMs
# Yarn:NM丢失节点数量统计
Hadoop_ResourceManager_NumLostNMs
# Yarn:NM不健康节点数量统计
Hadoop_ResourceManager_NumUnhealthyNMs
# Yarn:app运行数量
sum(Hadoop_ResourceManager_AppsRunning{q1 != ""})
# Yarn:app提交数量
sum(Hadoop_ResourceManager_AppsSubmitted{q1 != ""})
# Yarn:app等待数量
sum(Hadoop_ResourceManager_AppsPending{q1 != ""})
# Yarn:app完成数量
sum(Hadoop_ResourceManager_AppsCompleted{q1 != ""})
# Yarn:app被kill的数量
sum(Hadoop_ResourceManager_AppsKilled{q1 != ""})
# Yarn:app失败数量
sum(Hadoop_ResourceManager_AppsFailed{q1 != ""})
# Yarn:已分配的内存大小
sum(Hadoop_ResourceManager_AllocatedMB{q1 != ""})
# Yarn:已分配的核数量
sum(Hadoop_ResourceManager_AllocatedVCores{q1 != ""})
# Yarn:已分配的Container数量
sum(Hadoop_ResourceManager_AllocatedContainers{q1 != ""})
# Yarn:可用的内存大小(单位:MB)
sum(Hadoop_ResourceManager_AvailableMB{q1 != ""})
# Yarn:1min内NM心跳汇报次数
increase(Hadoop_ResourceManager_NodeUpdateNumOps[1m])
# Yarn:NM心跳汇报处理时间
Hadoop_ResourceManager_NodeUpdateAvgTime
- 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
3.Hive监测
# Hive:CPU 使用率
sum(java_lang_OperatingSystem_ProcessCpuLoad{job=~"hiveserver2|metastore"})*100
# Hive:JVM堆内存使用率
sum(jvm_memory_bytes_used{job=~"hiveserver2|metastore",area="heap"})/sum(jvm_memory_bytes_max{job=~"hiveserver2|metastore",area="heap"})*100
# Hive:JVM GC Time(单位:ms)
increase(jvm_gc_collection_seconds_sum{job=~"hiveserver2|metastore"}[5m]) * 1000
# Hive:JVM GC Count(无单位)
increase(jvm_gc_collection_seconds_count{job=~"hiveserver2|metastore"}[5m])
# Hive:jvm当前线程数量(无单位)
sum(jvm_threads_current{job=~"hiveserver2|metastore"})
# Hive:jvm后台线程数量(无单位)
sum(jvm_threads_daemon{job=~"hiveserver2|metastore"})
# Hive:jvm死锁线程数量(无单位)
sum(jvm_threads_deadlocked{job=~"hiveserver2|metastore"})
# Hive:jvm线程阻塞数量(无单位)
sum(jvm_threads_state{job=~"hiveserver2|metastore",state="BLOCKED"})
# Hive:jvm线程等待数量(无单位)
sum(jvm_threads_state{job=~"hiveserver2|metastore",state="WAITING"})
# Hive:当前jvm已加载类数量(无单位)
sum(jvm_classes_loaded_total{job=~"hiveserver2|metastore"})
# Hive:用户和系统的总cpu使用时间(单位:s)
max(process_cpu_seconds_total{job=~"hiveserver2|metastore"})
- 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
4.Kafka监测
# Kafka:CPU使用率(%)
irate(process_cpu_seconds_total{job="kafka"}[5m])*100
# Kafka:JVM内存使用情况(单位:MiB)
sum without(area)(jvm_memory_bytes_used{job="kafka"}) / 1024 / 1024
# Kafka:Broker网络吞吐量-流入(单位:B/s)
sum(rate(kafka_server_brokertopicmetrics_bytesinpersec{job="kafka",topic!=""}[5m]))
# Kafka:Broker网络吞吐量-流出(单位:B/s)
sum(rate(kafka_server_brokertopicmetrics_bytesoutpersec{job="kafka",topic!=""}[5m]))
# Kafka:激活状态控制器数量(无单位)
sum(kafka_controller_kafkacontroller_activecontrollercount{job="kafka"})
# Kafka:在线分区数(无单位)
sum(kafka_server_replicamanager_partitioncount{job="kafka"})
# Kafka:离线分区数(无单位)
sum(kafka_controller_kafkacontroller_offlinepartitionscount{job="kafka"})
# Kafka:在线Broker数量(无单位)
count(kafka_server_replicamanager_leadercount{job="kafka"})
# Kafka:Broker消息速率(io/s)
sum without(topic)(rate(kafka_server_brokertopicmetrics_messagesinpersec{job="kafka"}[5m]))
# Kafka:Broker入站流量速率(B/s)
sum without(topic)(rate(kafka_server_brokertopicmetrics_bytesinpersec{job="kafka"}[5m]))
# Kafka:Broker出站流量速率(B/s)
sum without(topic)(rate(kafka_server_brokertopicmetrics_bytesoutpersec{job="kafka"}[5m]))
# Kafka:topic消息速率(io/s)
sum without(instance)(rate(kafka_server_brokertopicmetrics_messagesinpersec{job="kafka",topic!=""}[5m]))
# Kafka:topic入站流量速率(B/s)
sum without(instance)(rate(kafka_server_brokertopicmetrics_bytesinpersec{job="kafka",topic!=""}[5m]))
# Kafka:topic出站流量速率(B/s)
sum without(instance)(rate(kafka_server_brokertopicmetrics_bytesoutpersec{job="kafka",topic!=""}[5m]))
# Kafka:I-O线程池平均空闲比例(%)
kafka_server_kafkarequesthandlerpool_requesthandleravgidlepercent_total{job="kafka"}
# Kafka:活跃连接数
sum(kafka_server_socketservermetrics_connection_count{job="kafka"}) by (instance, listener)
- 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
5.Zookeeper监测
# Zookeeper:Quorum Size(无单位)
zookeeper_QuorumSize{job="zookeeper"}
# Zookeeper:Follower个数(无单位)
avg(zookeeper_QuorumSize{job="zookeeper"}) -1
# Zookeeper:健康检查(Ticktime)(单位:ms)
avg(zookeeper_TickTime{job="zookeeper"}) by (instance)
# Zookeeper:平均请求延迟
zookeeper_AvgRequestLatency{job="zookeeper"}
# Zookeeper:Znode个数
zookeeper_InMemoryDataTree_NodeCount{job="zookeeper"}
# Zookeeper:活跃连接数
zookeeper_NumAliveConnections{job="zookeeper"}
# Zookeeper:未完成请求数
zookeeper_OutstandingRequests{job="zookeeper"}
# Zookeeper:监视器数量
zookeeper_InMemoryDataTree_WatchCount{job="zookeeper"}
# Zookeeper:JVM GC Time(单位:ms)
increase(jvm_gc_collection_seconds_sum{job="zookeeper"}[5m]) * 1000
# Zookeeper:JVM GC Count(无单位)
increase(jvm_gc_collection_seconds_count{job="zookeeper"}[5m])
# Zookeeper:JVM堆内存使用率(%)
jvm_memory_bytes_used{job="zookeeper", area="heap"}/jvm_memory_bytes_max{job="zookeeper", area="heap"}*100
- 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