check_hadoop_balance.pl (Advanced Nagios Plugins Collection)

Project Notes

Part of the Advanced Nagios Plugins Collection, download it here:

https://github.com/harisekhon/nagios-plugins

./check_hadoop_balance.pl --help

Nagios Hadoop Plugin to check various health aspects of HDFS via the Namenode's dfsadmin -report

- checks % HDFS space used. Based off an earlier plugin I wrote in 2010 that we used in production for over 2 years. This heavily leverages HariSekhonUtils so code in this file is very short but still much tighter validated
- checks HDFS replication of blocks, again based off another plugin I wrote in 2010 around the same time as above and ran in production for 2 years. This code unifies/dedupes and improves on both those plugins
- checks HDFS % Used Balance is within thresholds
- checks number of available datanodes and if there are any dead datanodes

Originally written for old vanilla Apache Hadoop 0.20.x, updated for CDH 4.3 (2.0.0-cdh4.3.0)

Recommend you also investigate check_hadoop_cloudera_manager_metrics.pl (disclaimer I work for Cloudera but seriously it's good it gives you access to a wealth of information)

usage: check_hadoop_balance.pl [ options ]

-s --hdfs-space Checks % HDFS Space used against given warning/critical thresholds
-r --replication Checks replication state: under replicated blocks, corrupt blocks, missing blocks. Warning/critical thresholds apply to under replicated blocks. Corrupt and missing blocks if any raise critical since this means there is potentially data loss
-b --balance Checks Balance of HDFS Space used % across datanodes is within thresholds. Lists the nodes out of balance in verbose mode
-n --nodes-available Checks the number of available datanodes against the given warning/critical thresholds as the lower limits (inclusive). Any dead datanodes raises warning
-w --warning Warning threshold or ran:ge (inclusive)
-c --critical Critical threshold or ran:ge (inclusive)
--hadoop-bin Path to 'hdfs' or 'hadoop' command if not in $PATH
--hadoop-user Checks that this plugin is being run by the hadoop user (defaults to 'hdfs', falls back to trying 'hadoop' unless specified)
-h --help Print description and usage options
-t --timeout Timeout in secs (default: 10)
-v --verbose Verbose mode
-V --version Print version and exit