1011d9202c
- Fix permission denied on log files (chown nut:nut) - Fix upssched.conf permissions (root:nut) - Add sudo for perl to allow PVE::Notify from user nut - Add periodic battery status emails every minute when on battery - Add charging status emails at 5, 10, 30 min after power restore - Remove diacritics from all notification messages - Update documentation with sudo and permissions setup Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
437 lines
14 KiB
Bash
437 lines
14 KiB
Bash
#!/bin/bash
|
|
#
|
|
# Script apelat de upssched pentru a gestiona evenimentele UPS
|
|
# Trimite notificari email via PVE::Notify
|
|
#
|
|
# Creat: 2025-10-06
|
|
# Actualizat: 2026-01-14 - Adaugat notificari periodice status baterie
|
|
|
|
LOGFILE=/var/log/ups-events.log
|
|
UPS_NAME="nutdev1"
|
|
TEMPLATE_DIR="/etc/pve/notification-templates/default"
|
|
HOSTNAME=$(hostname)
|
|
FQDN=$(hostname -f 2>/dev/null || hostname)
|
|
BATTERY_MONITOR_PID="/run/nut/battery-monitor.pid"
|
|
|
|
# Asigura ca fisierul de log exista si are permisiunile corecte
|
|
# Scriptul ruleaza ca user 'nut', deci fisierul trebuie sa fie writable
|
|
ensure_logfile() {
|
|
if [ ! -f "$LOGFILE" ]; then
|
|
touch "$LOGFILE" 2>/dev/null || sudo touch "$LOGFILE"
|
|
fi
|
|
if [ ! -w "$LOGFILE" ]; then
|
|
sudo chown nut:nut "$LOGFILE" 2>/dev/null
|
|
sudo chmod 664 "$LOGFILE" 2>/dev/null
|
|
fi
|
|
}
|
|
|
|
ensure_logfile
|
|
|
|
log_event() {
|
|
echo "[$(date '+%Y-%m-%d %H:%M:%S')] $1" >> $LOGFILE 2>/dev/null || \
|
|
logger -t upssched-cmd "LOG_FALLBACK: $1"
|
|
}
|
|
|
|
# Obtine status UPS
|
|
get_ups_status() {
|
|
upsc $UPS_NAME ups.status 2>/dev/null || echo "UNKNOWN"
|
|
}
|
|
|
|
get_battery_charge() {
|
|
upsc $UPS_NAME battery.charge 2>/dev/null || echo "0"
|
|
}
|
|
|
|
get_input_voltage() {
|
|
upsc $UPS_NAME input.voltage 2>/dev/null || echo "0"
|
|
}
|
|
|
|
get_battery_runtime() {
|
|
local runtime=$(upsc $UPS_NAME battery.runtime 2>/dev/null)
|
|
if [ -n "$runtime" ] && [ "$runtime" != "0" ]; then
|
|
echo $((runtime / 60))
|
|
else
|
|
echo ""
|
|
fi
|
|
}
|
|
|
|
get_battery_voltage() {
|
|
upsc $UPS_NAME battery.voltage 2>/dev/null || echo "0"
|
|
}
|
|
|
|
get_ups_load() {
|
|
upsc $UPS_NAME ups.load 2>/dev/null || echo "0"
|
|
}
|
|
|
|
# Creeaza template-uri daca nu exista
|
|
create_templates() {
|
|
mkdir -p $TEMPLATE_DIR
|
|
|
|
cat > "$TEMPLATE_DIR/ups-power-event-subject.txt.hbs" << 'EOFTEMPLATE'
|
|
[{{ hostname }}] UPS {{ event_type }} - {{ event_title }}
|
|
EOFTEMPLATE
|
|
|
|
cat > "$TEMPLATE_DIR/ups-power-event-body.txt.hbs" << 'EOFTEMPLATE'
|
|
========================================
|
|
UPS POWER EVENT - {{ event_title }}
|
|
========================================
|
|
|
|
Hostname: {{ hostname }}
|
|
Date: {{ event_date }}
|
|
Event: {{ event_type }}
|
|
|
|
{{ event_description }}
|
|
|
|
UPS STATUS:
|
|
-----------
|
|
Status: {{ ups_status }}
|
|
Battery Charge: {{ battery_charge }}%
|
|
Input Voltage: {{ input_voltage }}V
|
|
{{#if battery_runtime}}
|
|
Battery Runtime: {{ battery_runtime }} min
|
|
{{/if}}
|
|
|
|
{{#if action_taken}}
|
|
ACTION: {{ action_taken }}
|
|
{{/if}}
|
|
|
|
{{#if next_steps}}
|
|
NEXT STEPS: {{ next_steps }}
|
|
{{/if}}
|
|
|
|
========================================
|
|
Log: /var/log/ups-events.log
|
|
========================================
|
|
EOFTEMPLATE
|
|
|
|
cat > "$TEMPLATE_DIR/ups-power-event-body.html.hbs" << 'EOFTEMPLATE'
|
|
<!DOCTYPE html>
|
|
<html>
|
|
<head>
|
|
<meta charset="UTF-8">
|
|
<style>
|
|
body { font-family: Arial, sans-serif; margin: 0; padding: 20px; background-color: #f5f5f5; }
|
|
.container { max-width: 600px; margin: 0 auto; background: white; padding: 25px; border-radius: 8px; box-shadow: 0 2px 8px rgba(0,0,0,0.1); }
|
|
h1 { margin-top: 0; padding-bottom: 15px; border-bottom: 3px solid #3498db; }
|
|
.status-onbatt { color: #e67e22; border-color: #e67e22; }
|
|
.status-online { color: #27ae60; border-color: #27ae60; }
|
|
.status-lowbatt { color: #c0392b; border-color: #c0392b; }
|
|
.status-shutdown { color: #8e44ad; border-color: #8e44ad; }
|
|
.status-charging { color: #3498db; border-color: #3498db; }
|
|
.alert-box { padding: 15px; border-radius: 5px; margin: 20px 0; border-left: 4px solid; }
|
|
.alert-warning { background: #fef9e7; border-color: #f39c12; }
|
|
.alert-danger { background: #fdedec; border-color: #e74c3c; }
|
|
.alert-success { background: #eafaf1; border-color: #2ecc71; }
|
|
.alert-info { background: #ebf5fb; border-color: #3498db; }
|
|
.metrics { display: grid; grid-template-columns: repeat(2, 1fr); gap: 10px; margin: 20px 0; }
|
|
.metric { background: #ecf0f1; padding: 15px; border-radius: 5px; text-align: center; }
|
|
.metric-label { font-size: 12px; color: #7f8c8d; text-transform: uppercase; }
|
|
.metric-value { font-size: 24px; font-weight: bold; color: #2c3e50; margin-top: 5px; }
|
|
.footer { margin-top: 25px; padding-top: 15px; border-top: 1px solid #ecf0f1; font-size: 12px; color: #7f8c8d; }
|
|
</style>
|
|
</head>
|
|
<body>
|
|
<div class="container">
|
|
<h1 class="status-{{ event_class }}">[UPS] {{ event_title }}</h1>
|
|
<p><strong>Hostname:</strong> {{ hostname }}<br>
|
|
<strong>Date:</strong> {{ event_date }}<br>
|
|
<strong>Event:</strong> {{ event_type }}</p>
|
|
<div class="alert-box alert-{{ alert_type }}">
|
|
<strong>{{ event_description }}</strong>
|
|
</div>
|
|
<h3>UPS Status</h3>
|
|
<div class="metrics">
|
|
<div class="metric">
|
|
<div class="metric-label">Status</div>
|
|
<div class="metric-value">{{ ups_status }}</div>
|
|
</div>
|
|
<div class="metric">
|
|
<div class="metric-label">Battery</div>
|
|
<div class="metric-value">{{ battery_charge }}%</div>
|
|
</div>
|
|
<div class="metric">
|
|
<div class="metric-label">Input Voltage</div>
|
|
<div class="metric-value">{{ input_voltage }}V</div>
|
|
</div>
|
|
</div>
|
|
{{#if action_taken}}
|
|
<p><strong>Action:</strong> {{ action_taken }}</p>
|
|
{{/if}}
|
|
{{#if next_steps}}
|
|
<p><strong>Next Steps:</strong> {{ next_steps }}</p>
|
|
{{/if}}
|
|
<div class="footer">
|
|
<p>Log: /var/log/ups-events.log<br>Proxmox UPS Monitoring</p>
|
|
</div>
|
|
</div>
|
|
</body>
|
|
</html>
|
|
EOFTEMPLATE
|
|
|
|
log_event "Templates created in $TEMPLATE_DIR/"
|
|
}
|
|
|
|
# Trimite notificare email via PVE::Notify
|
|
send_notification() {
|
|
local EVENT_TYPE="$1"
|
|
local EVENT_TITLE="$2"
|
|
local EVENT_DESC="$3"
|
|
local EVENT_CLASS="$4"
|
|
local ALERT_TYPE="$5"
|
|
local SEVERITY="$6"
|
|
local ACTION="$7"
|
|
local NEXT_STEPS="$8"
|
|
|
|
# Verifica si creeaza template-uri
|
|
if [ ! -f "$TEMPLATE_DIR/ups-power-event-subject.txt.hbs" ]; then
|
|
create_templates
|
|
fi
|
|
|
|
# Obtine status UPS
|
|
local UPS_STATUS=$(get_ups_status)
|
|
local BATTERY_CHARGE=$(get_battery_charge)
|
|
local INPUT_VOLTAGE=$(get_input_voltage)
|
|
local BATTERY_RUNTIME=$(get_battery_runtime)
|
|
local EVENT_DATE=$(date '+%Y-%m-%d %H:%M:%S')
|
|
|
|
log_event "Sending $SEVERITY notification: $EVENT_TITLE"
|
|
|
|
/usr/bin/sudo /usr/bin/perl -I/usr/share/perl5 << EOFPERL 2>&1 | tee -a $LOGFILE
|
|
use strict;
|
|
use warnings;
|
|
use PVE::Notify;
|
|
|
|
my \$template_data = {
|
|
'hostname' => '$FQDN',
|
|
'event_date' => '$EVENT_DATE',
|
|
'event_type' => '$EVENT_TYPE',
|
|
'event_title' => '$EVENT_TITLE',
|
|
'event_description' => '$EVENT_DESC',
|
|
'event_class' => '$EVENT_CLASS',
|
|
'alert_type' => '$ALERT_TYPE',
|
|
'ups_status' => '$UPS_STATUS',
|
|
'battery_charge' => '$BATTERY_CHARGE',
|
|
'input_voltage' => '$INPUT_VOLTAGE',
|
|
'battery_runtime' => '$BATTERY_RUNTIME',
|
|
'action_taken' => '$ACTION',
|
|
'next_steps' => '$NEXT_STEPS'
|
|
};
|
|
|
|
my \$fields = {
|
|
'hostname' => '$HOSTNAME',
|
|
'type' => 'ups-power-event'
|
|
};
|
|
|
|
eval {
|
|
PVE::Notify::notify('$SEVERITY', 'ups-power-event', \$template_data, \$fields);
|
|
print "Email notification sent successfully\\n";
|
|
};
|
|
if (\$@) {
|
|
print STDERR "Failed to send notification: \$@\\n";
|
|
}
|
|
EOFPERL
|
|
}
|
|
|
|
# Porneste monitorizarea periodica a bateriei (la fiecare minut)
|
|
start_battery_monitor() {
|
|
# Opreste monitorul existent daca exista
|
|
stop_battery_monitor
|
|
|
|
log_event "Starting battery discharge monitor (every 60 seconds)"
|
|
|
|
# Porneste procesul de monitorizare in background
|
|
(
|
|
MINUTE=1
|
|
while true; do
|
|
sleep 60
|
|
|
|
# Verifica daca UPS-ul e inca pe baterie
|
|
STATUS=$(upsc $UPS_NAME ups.status 2>/dev/null)
|
|
if [[ ! "$STATUS" =~ "OB" ]]; then
|
|
log_event "Battery monitor: UPS no longer on battery, stopping monitor"
|
|
break
|
|
fi
|
|
|
|
CHARGE=$(get_battery_charge)
|
|
VOLTAGE=$(get_battery_voltage)
|
|
RUNTIME=$(get_battery_runtime)
|
|
LOAD=$(get_ups_load)
|
|
|
|
log_event "Battery monitor: Minute $MINUTE - Charge: ${CHARGE}%, Voltage: ${VOLTAGE}V, Load: ${LOAD}%"
|
|
|
|
send_notification \
|
|
"BATTERY_STATUS" \
|
|
"Status baterie - ${CHARGE}% (min $MINUTE)" \
|
|
"UPS pe baterie de $MINUTE minute. Baterie: ${CHARGE}%, Voltaj: ${VOLTAGE}V, Load: ${LOAD}%${RUNTIME:+, Runtime: ${RUNTIME} min}" \
|
|
"onbatt" \
|
|
"warning" \
|
|
"warning" \
|
|
"Monitorizare descarcare" \
|
|
"Se trimite status la fiecare minut"
|
|
|
|
MINUTE=$((MINUTE + 1))
|
|
done
|
|
rm -f $BATTERY_MONITOR_PID
|
|
) &
|
|
|
|
echo $! > $BATTERY_MONITOR_PID
|
|
log_event "Battery monitor started with PID $(cat $BATTERY_MONITOR_PID)"
|
|
}
|
|
|
|
# Opreste monitorizarea periodica a bateriei
|
|
stop_battery_monitor() {
|
|
if [ -f "$BATTERY_MONITOR_PID" ]; then
|
|
PID=$(cat $BATTERY_MONITOR_PID)
|
|
if [ -n "$PID" ] && kill -0 $PID 2>/dev/null; then
|
|
log_event "Stopping battery monitor (PID $PID)"
|
|
kill $PID 2>/dev/null
|
|
# Asteapta putin si forteaza daca nu s-a oprit
|
|
sleep 1
|
|
kill -9 $PID 2>/dev/null
|
|
fi
|
|
rm -f $BATTERY_MONITOR_PID
|
|
fi
|
|
}
|
|
|
|
# Trimite status incarcare baterie
|
|
send_charge_status() {
|
|
local MINUTES="$1"
|
|
local CHARGE=$(get_battery_charge)
|
|
local VOLTAGE=$(get_battery_voltage)
|
|
local INPUT=$(get_input_voltage)
|
|
local LOAD=$(get_ups_load)
|
|
|
|
log_event "Charge status at $MINUTES min: ${CHARGE}%, Voltage: ${VOLTAGE}V, Input: ${INPUT}V"
|
|
|
|
send_notification \
|
|
"CHARGING_STATUS" \
|
|
"Incarcare baterie - ${CHARGE}% (+${MINUTES} min)" \
|
|
"Status incarcare la $MINUTES minute dupa revenirea curentului. Baterie: ${CHARGE}%, Voltaj baterie: ${VOLTAGE}V, Tensiune intrare: ${INPUT}V, Load: ${LOAD}%" \
|
|
"charging" \
|
|
"info" \
|
|
"info" \
|
|
"Monitorizare incarcare" \
|
|
"Bateria se incarca"
|
|
}
|
|
|
|
# Handler principal
|
|
case $1 in
|
|
onbatt_start)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Trecere pe baterie - Timer 3 minute pornit"
|
|
logger -t upssched-cmd "UPS switched to battery - 3 minute timer started"
|
|
|
|
send_notification \
|
|
"ONBATT" \
|
|
"Trecere pe baterie" \
|
|
"UPS a trecut pe baterie! Daca curentul nu revine in 3 minute, se va initia shutdown-ul cluster-ului." \
|
|
"onbatt" \
|
|
"warning" \
|
|
"warning" \
|
|
"Timer 3 minute pornit" \
|
|
"Asteptare revenire curent sau shutdown automat"
|
|
|
|
# Porneste monitorizarea periodica a bateriei
|
|
start_battery_monitor
|
|
;;
|
|
|
|
onbatt)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Pe baterie de 3 minute - Incepe shutdown orchestrat"
|
|
logger -t upssched-cmd "UPS on battery for 3 minutes - starting orchestrated shutdown"
|
|
|
|
# Opreste monitorizarea - urmeaza shutdown
|
|
stop_battery_monitor
|
|
|
|
send_notification \
|
|
"ONBATT_TIMEOUT" \
|
|
"Pe baterie 3 min - SHUTDOWN" \
|
|
"UPS a fost pe baterie timp de 3 minute. Se initiaza shutdown-ul orchestrat al cluster-ului." \
|
|
"onbatt" \
|
|
"danger" \
|
|
"error" \
|
|
"Pornire shutdown orchestrat cluster" \
|
|
"Toate nodurile se vor opri in ordine"
|
|
|
|
/usr/local/bin/ups-shutdown-cluster.sh &
|
|
;;
|
|
|
|
online)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Curent revenit - UPS online"
|
|
logger -t upssched-cmd "Power restored - UPS back online"
|
|
|
|
# Opreste monitorizarea descarcarii
|
|
stop_battery_monitor
|
|
|
|
send_notification \
|
|
"ONLINE" \
|
|
"Curent revenit - OK" \
|
|
"Curentul electric a revenit. UPS functioneaza normal pe linia AC. Se vor trimite statusuri de incarcare la 5, 10 si 30 de minute." \
|
|
"online" \
|
|
"success" \
|
|
"info" \
|
|
"Sistem stabil" \
|
|
"Monitorizare incarcare activa"
|
|
;;
|
|
|
|
charge_5min)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Status incarcare la 5 minute"
|
|
send_charge_status "5"
|
|
;;
|
|
|
|
charge_10min)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Status incarcare la 10 minute"
|
|
send_charge_status "10"
|
|
;;
|
|
|
|
charge_30min)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Status incarcare la 30 minute"
|
|
send_charge_status "30"
|
|
;;
|
|
|
|
lowbatt)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: BATERIE SCAZUTA - Shutdown IMEDIAT"
|
|
logger -t upssched-cmd "UPS LOW BATTERY - immediate shutdown"
|
|
|
|
# Opreste monitorizarea
|
|
stop_battery_monitor
|
|
|
|
send_notification \
|
|
"LOWBATT" \
|
|
"BATERIE CRITICA - SHUTDOWN IMEDIAT" \
|
|
"UPS raporteaza baterie critica! Shutdown imediat pentru protectia datelor." \
|
|
"lowbatt" \
|
|
"danger" \
|
|
"error" \
|
|
"Shutdown imediat cluster" \
|
|
"Sistemele se opresc ACUM"
|
|
|
|
/usr/local/bin/ups-shutdown-cluster.sh &
|
|
;;
|
|
|
|
commbad)
|
|
log_event "=========================================="
|
|
log_event "UPS EVENT: Comunicatie pierduta cu UPS de 30 secunde"
|
|
logger -t upssched-cmd "Lost communication with UPS for 30 seconds"
|
|
|
|
send_notification \
|
|
"COMMBAD" \
|
|
"Comunicatie pierduta cu UPS" \
|
|
"Nu se poate comunica cu UPS-ul de 30 de secunde. Verificati conexiunea USB." \
|
|
"commbad" \
|
|
"warning" \
|
|
"warning" \
|
|
"Monitorizare activa" \
|
|
"Verificati conexiunea fizica USB a UPS-ului"
|
|
;;
|
|
|
|
*)
|
|
log_event "UPS EVENT: Eveniment necunoscut - $1"
|
|
logger -t upssched-cmd "Unknown UPS event: $1"
|
|
;;
|
|
esac
|