Disable Resource Manager HA

1. Stop YARN and Zookeeper services from ambari
2. From Ambari-server :

                                           /var/lib/ambari-server/resources/scripts/configs.py --user=admin --     
                                                             password=admin --host=edgenode.hdp.cn --cluster=hdpdev --
                                                             action=get --config-      type=yarn-site -f yarn-site.json

3. From the below change the first property to value "false" and remove the other properties from yarn-site.json

                                       a) "yarn.resourcemanager.ha.enabled": "false",

                                  b) "yarn.resourcemanager.ha.rm-ids": "rm1,rm2",

                                  c) "yarn.resourcemanager.hostname.rm1": "datanode1.hdp.cn",

                                  d) "yarn.resourcemanager.hostname.rm2": "edgenode.hdp.cn",

                                  e)  "yarn.resourcemanager.webapp.address.rm1": "datanode1.hdp.cn:8088",

                                  f)  "yarn.resourcemanager.webapp.address.rm2": "edgenode.hdp.cn:8088",

                                 g) "yarn.resourcemanager.webapp.https.address.rm1": "datanode1.hdp.cn:8090",

                                 h) "yarn.resourcemanager.webapp.https.address.rm2": "edgenode.hdp.cn:8090",

                                 i) "yarn.resourcemanager.cluster-id": "yarn-cluster",

                                 j) "yarn.resourcemanager.ha.automatic-failover.zk-base-path": "/yarn-leader-election",

4. Set the below properties to existing resource manager

                                   a)  "yarn.resourcemanager.hostname":

                               b) "yarn.resourcemanager.admin.address":

                               c) "yarn.resourcemanager.webapp.address":                                                             

                               d)"yarn.resourcemanager.resource-tracker.address":                 

                               e)"yarn.resourcemanager.scheduler.address":

                                f)"yarn.resourcemanager.webapp.https.address":

                                g)"yarn.timeline-service.webapp.address":

                                h)"yarn.timeline-service.webapp.https.address":

                                 i)"yarn.timeline-service.address":

                                 j)"yarn.log.server.url":

5. Copy the yarn-site.json file back to the ambari-server and run the below command to set the changes made

                                            /var/lib/ambari-server/resources/scripts/configs.py --user=admin --password=admin --                                           host=edgenode.hdp.cn --cluster=hdpdev --action=get --config-type=yarn-site

6. Delete the Resource Manager host comonent

                                         curl --user admin:admin -i -H "X-Requested-By: ambari" -X  DELETE  
                                    http://edgenode.hdp.cn:8080/api/v1/clusters/hdpdev/hosts/edgenode.hdp.cn/
                                    host_components/RESOURCEMANAGER
                                                                                                                                                                                                        

7. Start zookeeper service from Ambari.

8. On  zookeeper clients run the below command to change the znode permissions

                                       /usr/hdp/current/zookeeper-client/bin/zkCli.sh getAcl /rmstore/ZKRMStateRoot

                                 /usr/hdp/current/zookeeper-client/bin/zkCli.sh setAcl /rmstore/ZKRMStateRoot                                                       world:anyone:rwcda

9. From Ambari UI, restart zookeeper service and YARN service

Yarn Capacity Scheduler

1. /etc/hadoop/conf/yarn-site.xml                                                                                                                                                                                    yarn.resourcemanager.scheduler.class=org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler

2. Capacity Scheduler settings (/etc/hadoop/conf/capacity-scheduler.xml on resource manager host)

  

   a) Queues are fundamental unit of scheduling in yarn

               yarn.scheduler.capacity.root.queues = support,engineering,marketing     (parent queues)

                       yarn.scheduler.capacity.support.queues=training,services                (leaf queues)

                       yarn.scheduler.capacity.engineering=development,qa                       (leaf queues)

                       yarn.scheduler.capacity.marketing=sales,advertising                         (leaf queues)

 

             b) Capacity Scheduler ACL's

                        

                         Queue users:

                         yarn.scheduler.capacity.root.acl_submit_application=      (asterisk == all users and groups & empty space == deny all users and groups)

                                  yarn.scheduler.capacity.support.acl_submit_applications= sherlock,pacioli,njella hg_group,cd_group

                                  yarn.scheduler.capacity.engineering.acl_submit_applications=user1,user2,user3 group1,group2,group3

                                  yarn.scheduler.capacity.marketing.acl_submit_applications= user4,user2,user5, group4,group5,group6

               Queue Administers

                         yarn.scheduler.capacity.root.acl_administer_queue=

                                  yarn.scheduler.capacity.support.acl_administer_queue= support_admin_group

                                  yarn.scheduler.capacity.engineering.acl_administer_queue=engineering_admin_group

                                  yarn.scheduler.capacity.marketing.acl_administer_queue=engineering_Admin_grp

   c) Queue Mappings:

               

               Setting Up Queue Mappings

                         yarn.scheduler.capacity.queue-mappings=u:njella:engineering,u:chitra:marketing,g:engineering_group:engineering

                         If group name is same as queue name then

                        yarn.scheduler.capacity.queue-mappings=u:%user:%primary_group                       

                        If user name is same as queue name then

                        yarn.scheduler.capacity.queue-mappings=u:%user:%user                       

               Enable queue mapping override:

                            yarn.scheduler.capacity.queue-mappings-override.enable=true

                         Example:                    

                                  <property>

                                          <name>yarn.scheduler.capacity.queue-mappings</name>

                                          <value>u:njella:engineering, g:webadmins:weblog</value> 

                                  </proprty>

                                            

                                            If user "njella" explicitly submits an application to the marketing queue, the default queue assignment of engineering is overriden and the  application is submitted to the marketing queue.

                     

Issues

1. Ambari-agent



Sol : Change verify=platform_default   to verify=disable  in /etc/pythoncert-verification.cfg file

1. sed -i 's/verify=platform_default/verify=disable/' /etc/python/cert-verification.cfg

-------------------------------------------------------------------------------------