Prerequisites

• HDP 2.0+
• YARN Installed
• At least 2 hosts in the cluster where YARN is not present on 1 host. These steps assume there are 3 hosts in the cluster.

Assumptions

version1

Installing ResourceManager

1. Stop all services except for HDFS. There are two methods that can be used to accomplish this:
   1. Using the Ambari web client to manually stop each service.
   2. Using the Ambari REST APIs directly. For this method, you can consult the documentation on starting and stopping services.
      
      

2. Add a ResourceManager component to YARN on a host that does not already have ResourceManager installed. This will not actually install ResourceManager, but will setup the host component associations.

   curl -u admin:$PASSWORD -H "X-Requested-By: Ambari" -i -X POST -d '{"host_components" : [{"HostRoles":{"component_name":"RESOURCEMANAGER"}}] }' http://<ambari-server>/api/v1/clusters/<cluster-name>/hosts?Hosts/host_name=<target-host>

3. Install ResourceManager on the same target_host used in the previous step.

   curl -u admin:$PASSWORD -H "X-Requested-By: Ambari" -i -X PUT -d '{"RequestInfo":{"context":"Install ResourceManager","operation_level":{"level":"HOST_COMPONENT","cluster_name":"<cluster-name>","host_name":"<target-host>","service_name":"YARN"}},"Body":{"HostRoles":{"state":"INSTALLED"}}}' http://<ambari-server>/api/v1/clusters/<cluster-name>/hosts/<target-host>/host_components/RESOURCEMANAGER

YARN Configuration

1. The yarn-site configuration group (yarn-site.xml) must now be updated to reflect the backup ResourceManager. 
   
   

   property	value
   yarn.resourcemanager.ha.enabled	true
   yarn.resourcemanager.ha.rm-ids	rm1,rm2
   yarn.resourcemanager.hostname.rm1	<rm-host-active>
   yarn.resourcemanager.hostname.rm2	<rm-host-standby>
   yarn.resourcemanager.recovery.enabled	true
   yarn.resourcemanager.store.class	org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore
   yarn.resourcemanager.zk-address	<zk-host-1>, <zk-host-2>, <zk-host-3>
   yarn.resourcemanager.cluster-id	<yarn-cluster-name>
   yarn.resourcemanager.ha.automatic-failover.zk-base-path	/yarn-leader-election

    

    

2. There are two methods to update these configuration properties:
   1. /var/lib/ambari-server/resources/scripts/configs.sh
      
   2. REST API
      
      

3. configs.sh

   $ ./configs.sh set <ambari-server> <cluster-name> yarn-site <param-name> <param-value>
    
   example
   $ ./configs.sh set c6401.ambari.apache.org cl1 yarn-site yarn.resourcemanager.ha.enabled true

   This will need to be execute for each parameter listed.

    

4. REST API

   1. Get the existing yarn-site configuration version and tag.

      GET http://<ambari-server>/api/v1/clusters/<cluster-name>?fields=Clusters/desired_configs
      ...
      {
        "href" : "http://<ambari-server>/api/v1/clusters/<cluster-name>?fields=Clusters/desired_configs",
        "Clusters" : {
          "cluster_name" : "c1",
          "version" : "HDP-2.1",
          "desired_configs" : {
            ...
            "yarn-site" : {
              "tag" : "version1",
              "user" : "_anonymous",
              "version" : 1
            },
            ...
          }
        }
      }

   2. Using the value from the tag property of yarn-site, request the current YARN configuration.

      GET http://<ambari-server>/api/v1/clusters/<cluster-name>/configurations?(type=yarn-site&tag=<yarn-site-tag>)
      ...
        "href" : "http://<ambari-server>/api/v1/clusters/<cluster-name>/configurations?(type=yarn-site&tag=version1)",
        "items" : [
          {
            "href" : "http://<ambari-server>/api/v1/clusters/<cluster-name>/configurations?type=yarn-site&tag=version1",
            "tag" : "version1",
            "type" : "yarn-site",
            "version" : 1,
            "Config" : {
              "cluster_name" : "c1"
            },
            "properties" : {
              "yarn.acl.enable" : "false",
              "yarn.admin.acl" : "",
              ...
              "yarn.timeline-service.ttl-enable" : "true",
              "yarn.timeline-service.ttl-ms" : "2678400000",
              "yarn.timeline-service.webapp.address" : "c6402.ambari.apache.org:8188",
              "yarn.timeline-service.webapp.https.address" : "c6402.ambari.apache.org:8190"
            }
          }
        ]
      }

   3. The properties property contains the current yarn-site configuration. Append the new properties to this structure and then update yarn-site.

      PUT http://<ambari-server>/api/v1/clusters/<cluster-name>
      
      {  
         "Clusters":{  
            "desired_config":{  
               "type":"yarn-site",
               "tag":"<new-yarn-site-tag>",
               "properties"::{  
                  "yarn.acl.enable":"false",
                  "yarn.admin.acl":"",
                  ... 
                  "yarn.timeline-service.ttl-enable":"true",
                  "yarn.timeline-service.ttl-ms":"2678400000",
                  "yarn.timeline-service.webapp.address":"c6402.ambari.apache.org:8188",
                  "yarn.timeline-service.webapp.https.address":"c6402.ambari.apache.org:8190"
                  ...
                  <yarn-ha-site-properties>
                  "yarn.resourcemanager.ha.enabled": "true",
                  "yarn.resourcemanager.ha.rm-ids": "rm1,rm2",
      			...
               }
            }
         }
      }

5. Restart all stopped services.