check_unused_reservations

Check if a reserved node is idle

Note: it is assumed that TORQUE is used as a resource manager and Maui as a scheduler. Tested with TORQUE 2.3.5 and Maui 3.2.6p21.

Usage

check_unused_reservations -H host -w Wtime -c Ctime

e.g. check_multi -H 10.10.1.23 -w 3600 -c 86400 check reservations on node 10.10.1.23; state is WARNING if node has been in Idle state for 1 hour, CRITICAL if in Idle state for 1 day

#!/usr/bin/perl -w # nagios: -epn # doesn't work w/ ePN... # # pc 20090508 # if a node is reserved AND is idle for a long time ==> no good! use strict ; use lib "/opt/nagios/libexec" ; use utils qw($TIMEOUT %ERRORS &print_revision &support) ; use vars qw($PROGNAME) ; use Getopt::Long ; $PROGNAME = "check_unused_reservations" ; my $node ; my $warntime = 2 * 3600 ; my $crittime = 24 * 3600 ; GetOptions( "hostname|H:s" => \$node, "warn-time|w:i" => \$warntime, "crit-time|c:i" => \$crittime ) ; open CHECK, "/opt/nagios/libexec/external/checknode $node |" or exit $ERRORS{'UNKNOWN'} ; open LOG, ">>/tmp/unused_reservations.log" ; print LOG "starting...on node $node\n" ; my $idle_time_string ; while ( <CHECK> ) { # search for state line print LOG if $node eq "p001" ; chomp ; next unless m/^State:\s+([^ \t]+)\s+\(in current state for ([0-9:]+\d)\)/ ; if ( $1 eq "Idle" ) { $idle_time_string = $2 ; last ; } else { print "OK - not in idle state\n" ; exit $ERRORS{'OK'} ; } } $idle_time_string =~ m/ (?: (?: (?: (\d+): )? # optional d (\d{1,2}): # hours )? # optional d:hh (\d{1,2}): # minutes )? # optional d:hh:mm (\d{1,2}) # seconds /x ; my $days = ( $1 ? $1 : 0 ) ; my $hours = ( $2 ? $2 : 0 ) ; my $minutes = ( $3 ? $3 : 0 ) ; my $seconds = $4 ; my $total = $seconds + 60 * $minutes + 3600 * $hours + 86400 * $days ; while ( <CHECK> ) { # skip until reservations print LOG if $node eq "p001" ; chomp ; last if m/^Reservations:/ ; } my $active_reservation = 0 ; my $res_start_time = 0 ; while ( <CHECK> ) { # look for active user reservations print LOG if $node eq "p001" ; chomp ; next unless m/ ^\s*User\s+ '[^']+?' \(x\d+\)\s+ (-?(?:\d+|INFINITY)) /x ; $res_start_time = ( $1 ? $1 : 0 ) ; $active_reservation = 1 if $res_start_time =~ m/^-/ ; } if ( $active_reservation and ( $total > $crittime ) ) { print "CRITICAL - node $node is reserved and has been idle for $total seconds|idletime=$total\n" ; exit $ERRORS{'CRITICAL'} ; } if ( $active_reservation and ( $total > $warntime ) ) { print "WARNING - node $node is reserved and has been idle for $total seconds|idletime=$total\n" ; exit $ERRORS{'WARNING'} ; } my $res_str = ( $active_reservation ? '' : ' - no active reservation' ) ; print "OK$res_str|idletime=$total\n" ; exit $ERRORS{'OK'} ;
download code