We are running hortonworks cluster with 18 nodes with amazon ec2 instances. We can't guarantee the ec2 instances availability for a long time. Let us consider a case, two of my nodes dead in the cluster. What are the subsequent actions to be taken to add new instances. Do you guys tried hdp with amazon instances, have any procedure templates ? ... View more

We implemented ambari-metrics-collector in embedded mode. Now, it is taking more disk space under /hadoop/var/lib/ambari-metrics-collector/hbase/data/default. How can we delete prior data in the metrics collector? What are the precautions to be taken while deleting directories in the above location. Which directories need to be untouched? I saw Metric Aggregator took 2.8GB,Metric Record monitor took 1.4 GB. ... View more

We are running our cluster on CentOS 7. We have a plan to implement yarn new features such as CPU isolation. CPU isolation needs cgroups. May i implement cgroups without implementing Kerberos? ... View more

I created four node labels, assigned 4 nodes to each node label. Now, the jobs are running fine in the specified queues and on assigned nodes through node labels. Let us consider a case, two jobs are running in the same queue. Does one job which is acquiring more cluster resources will kill another job in the same queue? How can we isolate two jobs without impacting each other? Is there any solution? ... View more