Restarting and Maintaining Components

Last updated: 2020-07-29 15:24:14

    Feature Overview

    After a component configuration item is modified, the corresponding service needs to be restarted for the configuration to take effect. To minimize or eliminate the impact of service restart on the business continuity, the service can be restarted on a rolling basis. For instances in master/slave architecture, the slave instance will be restarted first before the master instance. Rolling restart takes more time than normal restart.

    • You can restart services in the console, and rolling restart is checked by default. Note: if rolling restart is disabled, all nodes will be restarted at the same time, and the services may become unavailable. Please do so with caution.
    • There are two policies for failure-handling when node restart fails: waiting for processing in case of failure or continuing processing in case of single node failure.

    Directions

    1. Log in to the EMR Console, select Cluster List, and click the ID/Name of the target cluster to enter the cluster details page.
    2. In Cluster Service, click Restart Service, or select the block of the component to be restarted, check the service role to be restarted in Operation > Role Management, and click Restart Service.
      • When restarting the service directly on the cluster details page, you need to select the service name, service role, restart method, whether to enable rolling restart, and failure-handling policy.
      • When restarting the service in the component block, you only need to select the restart method and failure-handling policy.
    3. The restart methods supported by the service components are as follows:
      Component Service Restart Mode Description Remarks
      HDFS NameNode Quick restart mode It can be restarted by running `hadoop-daemon.sh stop | start namenode` -
      NameNode Safe restart mode For an HA cluster, run `saveNameSpace` on the standby NameNode. Then, restart the component by running `hadoop-daemon.sh stop | start namenode` . For a non-HA cluster, the quick restart mode is used Only rolling restart is supported
      DataNode Default restart mode It can be restarted by running `hadoop-daemon.sh stop | start datanode` -
      JournalNode Default restart mode It can be restarted by running `hadoop-daemon.sh stop | start journalnode` -
      zkfc Default restart mode It can be restarted by running `hadoop-daemon.sh stop | start zkfc` -
      YARN ResourceManager Default restart mode It can be restarted by running `sbin/yarn-daemon.sh stop | start resourcemanager` Only rolling restart is supported
      NodeManager Default restart mode It can be restarted by running `sbin/yarn-daemon.sh stop | start nodemanager` -
      JobHisotryServer Default restart mode It can be restarted by running `sbin/yarn-daemon.sh stop| start historyserver` -
      TimeLineServer Default restart mode It can be restarted by running `sbin/yarn-daemon.sh stop | start timelineserver` -
      HBASE HbaseThrift Default restart mode It can be restarted by running `hbase-daemon.sh stop | start thrift` -
      HMaster Default restart mode It can be restarted by running `hbase-daemon.sh stop | start master` -
      HRegionServer Quick restart mode It can be restarted by running `hbase-daemon.sh stop | start regionserver` -
      HRegionServer Safe restart mode It can be restarted by running `graceful_stop.sh --restart --reload` -
      HIVE HiveMetaStore Default restart mode It can be restarted by running `hcat_server.sh stop | strat` -
      HiveServer2 Default restart mode It can be restarted by running `hive-daemon.sh stop-h2 | start-h2` -
      HiveWebHcat Default restart mode It can be restarted by running `webhcat_server.sh stop | start` -
      PRESTO PrestoCoordinator Default restart mode It can be restarted by running `bin/launcher stop | start` Only rolling restart is supported
      PrestoWorker Default restart mode It can be restarted by running `bin/launcher stop | start` -
      ZooKeeper QuorumPeerMain Default restart mode It can be restarted by running `bin/zkServer.sh stop | start` -
      Spark SparkJobHistoryServer Default restart mode It can be restarted by running `sbin/stop-history-server.sh | sbin/start-history-server.sh` -
      Hue Hue Default restart mode It can be restarted by running `build/env/bin/start.sh` and `build/env/bin/sop.sh` -
      Oozie Oozie Default restart mode It can be restarted by running `oozied.sh stop | start` -
      Storm Nimbus Default restart mode It can be restarted by running `bin/storm-daemon.sh nimbus stop | start` -
      Supervisor Default restart mode It can be restarted by running `bin/storm-daemon.sh supervisor stop | start` -
      LogViewer Default restart mode It can be restarted by running `bin/storm-daemon.sh nimbus stop | start` -
      UI Default restart mode It can be restarted by running `bin/storm-daemon.sh nimbus stop | start` -
      Ranger Ranger Default restart mode It can be restarted by running `sbin/ranger-daemon.sh stop | start` -
      Alluxio AlluxioMaster Default restart mode It can be restarted by running `bin/alluxio-stop.sh master` and `bin/alluxio-start.sh master` -
      AlluxioWorker Default restart mode It can be restarted by running `bin/alluxio-stop.sh worker` and `bin/alluxio-start.sh worker` -
      Ganglia Httpd Default restart mode It can be restarted by running `/sbin/service http stop | start` -
      gmetad Default restart mode It can be restarted by running `/sbin/service gmetad stop | start` -
      gmond Default restart mode It can be restarted by running `/sbin/service gmon stop | start` -

    Was this page helpful?

    Was this page helpful?

    • Not at all
    • Not very helpful
    • Somewhat helpful
    • Very helpful
    • Extremely helpful
    Send Feedback
    Help