Free CCA-500: Cloudera Certified Administrator for Apache Hadoop (CCAH) Exam Questions and Answers

On a CDH 5 cluster with all settings at their defaults, which port serves the NameNode web UI?

• AA. 8020
• BB. 8088
• CC. 50070
• DD. 50075

Your cluster runs MapReduce v2 on YARN with yarn.log-aggregation-enable set to true. A job finished an hour ago and you need to read the logs of its containers. Which two methods can you use? (Choose two.)

Select 2 answers.

• AA. Run yarn logs -applicationId with the application ID from a client machine
• BB. Use the Logs link on the NameNode web UI, which lists container logs for every application
• CC. Browse the job's task logs through the JobHistory Server web UI
• DD. Read the files in yarn.nodemanager.log-dirs on each worker node, which are retained after aggregation

An administrator adds the following lines to hadoop-metrics2.properties on every host and restarts all daemons. *.sink.file.class=org.apache.hadoop.metrics2.sink.FileSink *.period=10 namenode.sink.file.filename=namenode-metrics.out Metrics are written to namenode-metrics.out on the NameNode host, but no metrics file appears on the DataNode hosts. What is the most likely explanation?

• AA. The *.period property applies only to the NameNode, so DataNode metrics are never emitted
• BB. No datanode.sink.file.filename property is defined, so the FileSink for the DataNode writes to standard output
• CC. DataNode metrics are exposed only through JMX, so no sink can write them to a file
• DD. FileSink supports only NameNode and ResourceManager sources, so DataNodes require the GangliaSink

A user runs a MapReduce job with hadoop jar on a cluster running YARN. The client has copied the job JAR, configuration and input split metadata to HDFS and has submitted the application. What happens next, before any map task starts?

• AA. The NodeManager closest to the client starts a JobTracker process dedicated to the job
• BB. The ResourceManager allocates a container on a NodeManager, which launches the MapReduce ApplicationMaster there
• CC. The ResourceManager reads the input splits and starts a map task on each node holding a block of the input
• DD. The client contacts the NodeManagers directly and launches the ApplicationMaster on the node holding the first input block

Your company will deploy a 30-node production cluster running CDH 5. The standard server image is Windows Server 2012. The operations team can also support CentOS 6 and the most recent Fedora release. Which operating system plan is MOST appropriate?

• AA. Deploy Windows Server 2012 on every node and run the Hadoop daemons under Cygwin
• BB. Deploy the most recent Fedora release with the newest kernel on every node
• CC. Deploy CentOS 6 on every node of the cluster
• DD. Deploy Windows Server 2012 on the master nodes and CentOS 6 on the worker nodes

You restart the NameNode of a running cluster. The NameNode enters safe mode and stays there for several minutes. During this time, how does the NameNode learn which DataNodes hold the replicas of each block?

• AA. It reads the block locations from the edits log that it replays at startup
• BB. It receives block reports sent by each DataNode
• CC. It reads the block locations from the fsimage file that it loads at startup
• DD. It requests the block map from the Secondary NameNode's last checkpoint

The log4j.properties file for a Hadoop daemon contains the following lines: hadoop.root.logger=INFO,RFA log4j.appender.RFA=org.apache.log4j.RollingFileAppender log4j.appender.RFA.MaxFileSize=${hadoop.log.maxfilesize} What is the effect of these settings?

• AA. Events at INFO level and above go to a console appender that writes to stdout
• BB. Events at INFO level and above are sent over a socket to a remote log server
• CC. Events at INFO level and above go to a log file that starts a new daily file at midnight
• DD. Events at INFO level and above go to a log file that rolls when it reaches the configured maximum size

You installed Sqoop from CDH 5 packages on a gateway machine. An import from a MySQL database fails with the error: Could not load db driver class: com.mysql.jdbc.Driver. What should you do?

• AA. Copy the MySQL JDBC driver JAR into the Sqoop library directory on the gateway machine
• BB. Set the MySQL driver class in the Hive metastore configuration on the gateway machine
• CC. Copy the MySQL JDBC driver JAR into the lib directory of every DataNode in the cluster
• DD. Upload the MySQL JDBC driver JAR to the HDFS directory /user/sqoop on the cluster

An administrator wants a script to retrieve current NameNode metrics, such as CapacityUsed and NumLiveDataNodes, as JSON without changing any configuration. Which approach should the administrator use?

• AA. Run hdfs fsck / and parse the summary at the end of the report
• BB. Run hdfs dfsadmin -metasave and parse the output file
• CC. Send an HTTP request to the /jmx servlet on the NameNode web UI port
• DD. Parse the NameNode log file in /var/log/hadoop-hdfs

You add 10 new DataNodes to a cluster whose existing DataNodes are about 80% full. The new nodes are nearly empty. You want to move existing blocks so that the utilization of every DataNode is within 10 percentage points of the cluster average. Which command do you run?

• AA. hdfs fsck / -move
• BB. hdfs dfsadmin -refreshNodes
• CC. hdfs balancer -threshold 10
• DD. hdfs dfsadmin -setBalancerBandwidth 10485760