I added some new nodes to my cluster and it works fine. Then I add Spark Gateway roles to all the new nodes. We're using Yarn to manage and distribute Spark work. Does adding Spark Gateway roles to new nodes enough to make Yarn think like "Hey there are some new nodes here, let's distribute some containers and work to these new nodes"? Or do I have to add Yarn Gateway roles to these new nodes too? How to make sure that Yarn will use these new nodes when executing jobs to reduce the overall workload of my cluster     ... View more

I'm using CDH 6.2. I add and change some Yarn config inside Cloudera Manager Web UI (related to yarn shuffle service), something like "NodeManager Advanced Configuration Snippet (Safety Valve) for yarn-site.xml" I preview the result and Cloudera shows that some lines will be added to yarn-site.xml. But after restarting everything, I check yarn-site.xml at "/etc/hadoop/conf.cloudera.yarn/yarn-site.xml", I can't find any setting above, even the word "shuffle" doesn't appear anywhere at all? I also check for other files like marped-site.xml or core-site.xml, no luck so far     So where the heck do all the settings above add to? I'm confused. I need the proper yarn-site.xml file with all the settings just like in the Cloudera web UI     ... View more

I followed this guide on the Cloudera website, I've finished "Installing and Upgrading the Sentry Service" step. Now what do I do next? I tried to start Beeline and execute some queries as normal but get a privilege error kinit -k -t /home/vgdata/vega.keytab vega@BI.VEGA.COM  (get tgt for user vega) beeline -u "jdbc:hive2://data-node01:10000/test;principal=hive/data-node01.vega.com@BI.VEGA.COM" (access beeline with kerberos principal) select * from test; (execute query)   But I got an error like below Error: Error while compiling statement: FAILED: SemanticException No valid privileges User vega does not have privileges for SWITCHDATABASE The required privileges: Server=server1->Db=*->Table=+->Column=*... User "vega" is hdfs superuser, and also the main user that we use to connect to all Hadoop services (there is a Kerberos principal "vega" too) I tried to execute some commands like create role admin; grant role .... But all get an error like "No groups found for user vega"   Now where do I start? Is there a default "admin" user that I can do everything (including grant...)? I want to grant user "vega" the "admin" role, which can do everything. Something like my SQL grant privilege command GRANT ALL PRIVILEGES ON database.table TO user; GRANT ALL PRIVILEGES ON *.* TO vega;   ... View more