Database Reference
In-Depth Information
Figure 7-22.
Nagios root partition alert
This is an example of a monitored alert that could enable critical intervention when your Hadoop cluster
resources become limited. Monitoring and alerting on Hadoop cluster resources should be a mandatory
consideration when you are setting up a cluster yourself.
A combination of Naios and Ganglia could be used on the Hadoop cluster to provide a rich selection of historical
graphs and alerts. Nagios could trigger an alert for Hadoop cluster resources as those resources run low; then Ganglia
could examine the graphs that have recorded the condition of the resources over time.
Summary
This chapter discussed three major enhancements for Hadoop functionality. Hue provides a central location for
scripting and Hadoop-based job monitoring. Through Hue, you can inspect Hive and HBase databases and can
manipulate their data. Hue also offers the ability to visually browse the Hadoop file system HDFS.
Although Hue consolidates Hadoop fuctionality and access into one useful interface, it isn't a full-scale
monitoring system. (For instance, it doesn't have the ETL or reporting functionality of Pentaho or Talend, which
are covered in Chapters 10 and 11.) To supplement Hue, Ganglia and Nagios offer cluster monitoring. When used
together, Nagios and Ganglia complement each other: Nagios alerts the user to potential problems, while Ganglia
provides graph-based details to show what has happened, on which server and its type. This combined action will
help in problem investigation before system failure.
Search WWH ::
Custom Search