uptime - add new version with override feature & switch warning / critical value

defaults/main.yml

@@ -57,5 +57,5 @@ nrpe_proc_age_critical: 600
 # nrpe_redis_replication_lag_warning: 10
 # nrpe_redis_replication_lag_critical: 60
 
-nrpe_uptime_warning: 30
+nrpe_uptime_warning: 1440
-nrpe_uptime_critical: 1440
+nrpe_uptime_critical: 30

files/nrpe/check_uptime

@@ -6,16 +6,18 @@ STATE_WARNING=1
 STATE_CRITICAL=2
 STATE_UNKNOWN=3
 
-# Default values (in minutes)
+# Default values
 warn=1440
 crit=30
+OVERRIDE_FILE="/var/lib/nagios/uptime_override"
 
 # Retrieve arguments
-while getopts "w:c:" opt; do
+while getopts "w:c:f:" opt; do
   case $opt in
     w) warn=$OPTARG ;;
     c) crit=$OPTARG ;;
-    *) echo "Usage: $0 -w <minutes> -c <minutes>"; exit $STATE_UNKNOWN ;;
+    f) OVERRIDE_FILE=$OPTARG ;; # Permet de changer le chemin via Icinga
+    *) echo "Usage: $0 -w <minutes> -c <minutes> [-f <file_path>]"; exit $STATE_UNKNOWN ;;
   esac
 done
 
@@ -26,7 +28,18 @@ uptime_min=$((uptime_seconds / 60))
 # Performance Data
 PERFDATA="|uptime_min=${uptime_min}min;${warn};${crit};0;"
 
-# Decision logic
+# 1. Vérification du fichier d'acquittement
+if [ -f "$OVERRIDE_FILE" ]; then
+    # Si le système est redevenu "stable" naturellement, on supprime le flag
+    if [ "$uptime_min" -gt "$warn" ]; then
+        rm -f "$OVERRIDE_FILE"
+    else
+        echo "OK - Reboot acquitté manuellement (Uptime: ${uptime_min} min)$PERFDATA"
+        exit $STATE_OK
+    fi
+fi
+
+# 2. Logique de décision standard
 if [ "$uptime_min" -le "$crit" ]; then
     echo "CRITICAL - Recent reboot detected (${uptime_min} min < threshold ${crit} min)$PERFDATA"
     exit $STATE_CRITICAL
@@ -36,4 +49,4 @@ elif [ "$uptime_min" -le "$warn" ]; then
 else
     echo "OK - System stable (Uptime: ${uptime_min} min)$PERFDATA"
     exit $STATE_OK
-fi
+fi