Question-71: You are running cluster since last one year and using the Ambari to manage the same. Recently team is complaining that UI seems to be very slow. What you can do at first place?

Answer: As cluster is running since more than a year and you found that DB was never purged. So all the entire historical operations data is still in the database. Which most likely cause a performance issues in the UI. You need first delete or purge the history of the data, which you may not need. So you keep only 1 month recent history and delete all the other data. While purging the data it needs the cluster name as well, so if you are having more than one cluster monitored using the same Ambari server, then you can think of deleting the data for other cluster as well.

Question-72: For larger cluster where number of nodes are more than 1500, then what else you can think for Ambari optimization?

Answer: If there are more than 1500 nodes then we should do the tuning for the Ambari agent as well. Like it should not send entire status of each agent continuously, rather following needs to be done.

When on the agent if alert changes then only send the report to Ambari server else not. For example when agent status changes from OK to Warning then only send the status change information from agent to Server.

Question-73: What is the .pid file?

Answer: This is a process identification number of all the running process and stored in this file.

Question-74: What are the log file locations for the Ambari server and agent?

Answer: You can find log location at /var/log/ambari-agent for agent and /var/log/ambari-server for the Ambari-server.

Question-75: Is it possible to remove a particular host not be part of HDFS cluster operations like HDFS or YARN services?

Answer: Yes, we can do that. And that can be configured in the Ambari to manage which hosts in a cluster are included and excluded from participating in the cluster operations.