一、問題描述
Cdh的kafka本身有三節點，然後kafka也有三個broker，但是cdh1上的broker無論如何也無法啟動，報如下錯誤:Failed to restart service
二、問題原因
不知為何，進行探索
1.檢視日志
重新開機kafka，然後同時檢視kafka日志，好像沒報錯，沒發現什麼
[[email protected] ~]# tail -f /var/log/kafka/kafka-broker-cdh1.log
2020-02-12 04:59:56,077 INFO org.eclipse.jetty.util.log: Logging initialized @916ms
2020-02-12 04:59:56,229 WARN org.eclipse.jetty.server.handler.ContextHandler: [email protected]{/,null,null} contextPath ends with /*
2020-02-12 04:59:56,229 WARN org.eclipse.jetty.server.handler.ContextHandler: Empty contextPath
2020-02-12 04:59:56,715 INFO org.eclipse.jetty.server.Server: jetty-9.2.15.v20160210
2020-02-12 04:59:57,733 INFO org.eclipse.jetty.server.handler.ContextHandler: Started [email protected]{/,null,AVAILABLE}
2020-02-12 04:59:57,743 INFO org.eclipse.jetty.server.ServerConnector: Started [email protected]{HTTP/1.1}{0.0.0.0:24042}
2020-02-12 04:59:57,743 INFO org.eclipse.jetty.server.Server: Started @2583ms
2020-02-12 04:59:57,744 INFO nl.techop.kafka.KafkaHttpMetricsReporter: Started Kafka HTTP metrics reporter at http://0.0.0.0:24042/
2020-02-12 04:59:57,744 INFO kafka.server.KafkaServer: starting
2020-02-12 04:59:57,746 INFO kafka.server.KafkaServer: Connecting to zookeeper on cdh1:2181,cdh2:2181,cdh3:2181
2020-02-12 04:59:57,761 INFO org.I0Itec.zkclient.ZkEventThread: Starting ZkClient event thread.
2020-02-12 04:59:57,768 INFO org.apache.zookeeper.ZooKeeper: Client environment:zookeeper.version=3.4.10-39d3a4f269333c922ed3db283be479f9deacaa0f, built on 03/23/2017 10:13 GMT
2.轉看cm上顯示的報錯資訊
細節一看，有HeapDumpOnOutOfMemoryError 報錯，可能是超出堆記憶體大小了
再看報錯的最下面，也會有超出記憶體報錯java.lang.OutOfMemoryError，已經幾乎可以确認是由于堆記憶體大小不夠導緻
3.檢視kafka配置的堆記憶體大小，果然隻有50MB，幾乎幹不了事情，那肯定就是這個記憶體設定太小導緻的問題了。
三、解決問題
1.kafka配置broker的堆大小都修改為500MB
2.儲存重新開機後，即可
四、參考
https://blog.csdn.net/zhangxu1024/article/details/78852092