parse the dfsadmin report and give us some info about the cluster; warn if things are missing, if we have dead nodes, if we don't have enough datanodes, corrupt blocks, etc.

check_hadoop_dfs_report.pl

#!/usr/bin/perl
#
# This script is managed by puppet.
#

use strict;
use warnings;
use IO::File;
use Getopt::Long;

if (not exists $ENV{SUDO_USER} or $ENV{SUDO_USER} eq 'nagios') {
        exec("/usr/bin/sudo -u hdfs $0 @ARGV");
}

my $verbose;
my $help;

my @errors;
my %dfs_status;

my $hadoop_command = "/usr/bin/hadoop";
my $dfs_report     = "dfsadmin -report";

my $warn = "90";
my $crit_disk = "95";

my $avail_threshold = 80 / 100;

GetOptions(
        "help|h"        => \$help,
);

if (defined $help) {
        print <<EOF;
$0 [options]

        --help
        -h              This.

EOF

        exit 3;
}

my $fh = IO::File->new("$hadoop_command $dfs_report|");
if (not defined $fh) {
    print "UNKNOWN: could not run $hadoop_command $dfs_report: $!\n";
    exit 3;
}


while (<$fh>) {
    chomp;
    next if /^$/;
    next if /last contact/i;
    last if /^-----/; # break out to next section if we encounter the delimeter

    my ($key, $value) = split(/:/);
    $key=~s/\s+//g;

    if ($key =~ /dfsused%/i) {
        $value =~ s/%//;
    }
    $dfs_status{total}{lc $key} = $value;
}

while (<$fh>) {
    chomp;
    # Datanodes available: 12 (12 total, 0 dead)
    next unless /Datanodes available/;
    my ($avail_nodes, $total_nodes, $dead_nodes) = ($_ =~/Datanodes available:\s+(\d+)\s+\((\d+) total, (\d+) dead\)/);

    $dfs_status{nodes}{total} = $total_nodes;
    $dfs_status{nodes}{avail} = $avail_nodes;
    $dfs_status{nodes}{dead}  = $dead_nodes;
}


if (exists $dfs_status{total}) {
    #Key:Configured Capacity Value:  118481750196224 (107.76 TB)
    #Key:Present Capacity    Value:  118481750196224 (107.76 TB)
    #Key:DFS Remaining   Value:  46083857448971 (41.91 TB)
    #Key:DFS Used    Value:  72397892747253 (65.85 TB)
    #Key:DFS Used%   Value:  61.1%
    #Key:Under replicated blocks Value:  0
    #Key:Blocks with corrupt replicas    Value:  0
    #Key:Missing blocks  Value:  0

    if (exists $dfs_status{total}{'underreplicatedblocks'} and 
        defined $dfs_status{total}{'underreplicatedblocks'} and
                $dfs_status{total}{'underreplicatedblocks'} > 0) {
            push @errors, "Under replicated blocks > 0";
    }

    if (exists $dfs_status{total}{'blockswithcorruptreplicas'} and 
        defined $dfs_status{total}{'blockswithcorruptreplicas'} and
                $dfs_status{total}{'blockswithcorruptreplicas'} > 0) {
            push @errors, "Blocks with corrupt replicas > 0";
    }
    
    if (exists $dfs_status{total}{'missingblocks'} and 
        defined $dfs_status{total}{'missingblocks'} and
                $dfs_status{total}{'missingblocks'} > 0) {
            push @errors, "Missing blocks > 0";
    }

    if (exists $dfs_status{nodes}{'total'} and exists $dfs_status{nodes}{'avail'}) {
        if (($dfs_status{nodes}{'avail'} / $dfs_status{nodes}{'total'} ) < $avail_threshold) {
            push @errors, "Less than $avail_threshold % of datanodes available";
        }
    }

    if (exists $dfs_status{nodes}{'dead'} and $dfs_status{nodes}{'dead'} > 0) {
            push @errors, "Dead datanodes found in cluster";
    }

    if (exists $dfs_status{total}{'dfsused%'} ) {
        if (defined $dfs_status{total}{'dfsused%'} > $crit_disk) {
            push @errors, "DFS % space used $dfs_status{total}{'dfsused%'} > threshold $crit_disk %";
        }
    }
}



if (scalar @errors > 0) {
        print "CRITICAL: ", join("; ", @errors), "\n";
        exit 2;
} else {
        print "OK: dfsadmin report is happy.\n";
        exit 0;
}