When a problem occurs in a single node of a Hadoop cluster, it is generally not necessary to restart the entire system, just restart the node and it will automatically connect to the entire cluster.
Enter the following command on the necrotic node:
hadoop-daemon.sh Start Datanode
hadoop-daemon.sh Start Secondarynamenode
The cases are as follows:
Hadoop node crashes, can ping Pass, SSH Connection not on
Case:
Time: 2014/9/11 a.m.
Performance: tc-hadoop018 node on Hadoop web interface dead
Symptom: SSH connection to node tc-hadoop018 not on
Workaround:
Notify the computer room administrator to restart the machine,
Turn off firewall View status:/etc/init.d/iptatbles status closed:/etc/init.d/iptables stop
hadoop-daemon.sh Stop Datanode
hadoop-daemon.sh Stop Tasktracker
hadoop-daemon.sh Start Datanode
hadoop-daemon.sh Start Tasktracker
At this point, the startup succeeds
"Use caution" when necessary, view the Web interface and restart the entire cluster without running a job.
Hadoop the Secondarynamenode Port 50090 not through
Case:
Time: 2014/9/11 PM
Performance: Sos2 alarm prompt 123.125.244.6_50090 Port alarm
Symptom: JPS command 123.125.244.6 machine on Secondarynamenode process No
Workaround:
hadoop-daemon.sh Stop Secondarynamenode
hadoop-daemon.sh Start Secondarynamenode
At this point, the startup succeeds
Hadoop cluster Datanode Dead or Secondarynamenode process disappearance processing method