ajout du script check_smart
This commit is contained in:
@@ -37,4 +37,21 @@ for mount in "${MOUNTS[@]}"; do
|
||||
else
|
||||
$LOG_BIN INFO "disk_ok" "Disque $mount OK. $used_pct% utilisé."
|
||||
fi
|
||||
done
|
||||
done
|
||||
|
||||
# --- 2. Étude des Inodes ---
|
||||
# df -i récupère l'utilisation des inodes
|
||||
inode_pct="$(df -iP "$mount" 2>/dev/null | awk 'NR==2 {gsub("%","",$5); print $5}')"
|
||||
|
||||
if [[ ! "$inode_pct" =~ ^[0-9]+$ ]]; then
|
||||
$LOG_BIN ERROR "check_failed" "Erreur lecture inodes $mount."
|
||||
continue
|
||||
fi
|
||||
|
||||
if [ "$inode_pct" -ge "$CRITICAL" ]; then
|
||||
$LOG_BIN CRITICAL "inode_usage_critical" "Inodes $mount critiques ($inode_pct% utilisé)."
|
||||
elif [ "$inode_pct" -ge "$WARNING" ]; then
|
||||
$LOG_BIN WARNING "inode_usage_high" "Inodes $mount élevés ($inode_pct% utilisé)."
|
||||
else
|
||||
$LOG_BIN INFO "inode_ok" "Inodes $mount OK ($inode_pct% utilisé)."
|
||||
fi
|
||||
66
servers/linux/monitoring/bin/check_smart.sh
Normal file
66
servers/linux/monitoring/bin/check_smart.sh
Normal file
@@ -0,0 +1,66 @@
|
||||
#!/bin/bash
|
||||
#
|
||||
# This program is free software: you can redistribute it and/or modify
|
||||
# it under the terms of the GNU Affero General Public License as published by
|
||||
# the Free Software Foundation, either version 3 of the License, or
|
||||
# (at your option) any later version.
|
||||
#
|
||||
# This program is distributed in the hope that it will be useful,
|
||||
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
# GNU Affero General Public License for more details.
|
||||
|
||||
LOG_BIN="/opt/monitoring/bin/log-cli.php"
|
||||
|
||||
# --- Vérification ROOT ---
|
||||
if [ "${EUID}" -ne 0 ]; then
|
||||
echo "ERREUR : Ce script doit être exécuté en tant que root." >&2
|
||||
$LOG_BIN ERROR "internal_error" "Tentative d'exécution sans privilèges root."
|
||||
exit 1
|
||||
fi
|
||||
|
||||
# --- Vérification et installation de smartctl ---
|
||||
if ! command -v smartctl >/dev/null 2>&1; then
|
||||
# On tente l'installation (nécessite root, ce qui est le cas via cron)
|
||||
if command -v apt-get >/dev/null 2>&1; then
|
||||
apt-get update && apt-get install -y smartmontools
|
||||
fi
|
||||
|
||||
# Re-vérification après tentative
|
||||
if ! command -v smartctl >/dev/null 2>&1; then
|
||||
$LOG_BIN ERROR "internal_error" "smartctl non trouvé et installation impossible."
|
||||
exit 1
|
||||
fi
|
||||
fi
|
||||
|
||||
# On récupère les disques qui ont un transport physique (SATA, NVMe, USB)
|
||||
# Cela exclut d'office les /dev/mapper, /dev/dm-X, /dev/loopX
|
||||
DISKS=$(lsblk -dno NAME,TRAN | awk '$2!="" {print "/dev/"$1}')
|
||||
|
||||
for disk in $DISKS; do
|
||||
# Vérification : est-ce que smartctl peut lire ce périphérique ?
|
||||
# --scan-open vérifie si le disque est capable de répondre
|
||||
if ! smartctl -i "$disk" | grep -q "SMART support is: Enabled" 2>/dev/null; then
|
||||
# On peut logguer en INFO que le disque est ignoré car non-SMART (ex: clé USB basique)
|
||||
continue
|
||||
fi
|
||||
|
||||
# 1. État de santé global
|
||||
smart_output=$(smartctl -H "$disk" 2>/dev/null)
|
||||
exit_code=$?
|
||||
|
||||
if [ $exit_code -ne 0 ]; then
|
||||
$LOG_BIN CRITICAL "smart_health_bad" "État de santé PHYSIQUE CRITIQUE sur $disk"
|
||||
else
|
||||
# 2. Température
|
||||
temp=$(smartctl -A "$disk" 2>/dev/null | awk '/Temperature_Celsius/ {print $10}' | head -n 1)
|
||||
[ -z "$temp" ] && temp=$(smartctl -a "$disk" 2>/dev/null | awk '/Temperature:/ {print $2}' | head -n 1)
|
||||
|
||||
if [ -n "$temp" ]; then
|
||||
if [ "$temp" -ge 60 ]; then
|
||||
$LOG_BIN WARNING "disk_temp_high" "Surchauffe physique sur $disk : ${temp}°C"
|
||||
fi
|
||||
fi
|
||||
$LOG_BIN INFO "smart_health_ok" "Disque physique $disk sain."
|
||||
fi
|
||||
done
|
||||
@@ -140,6 +140,54 @@ function update_one_file($expected_hash, $mode, $rel_path) {
|
||||
return false;
|
||||
}
|
||||
|
||||
/**
|
||||
* Vérifie et ajoute les tâches cron si elles sont absentes
|
||||
*/
|
||||
function ensure_crontab_entries() {
|
||||
global $MONITORING_BASE_DIR;
|
||||
|
||||
// Définition des tâches souhaitées (Format: "cron_schedule command")
|
||||
$required_jobs = [
|
||||
"*/5 * * * * php {$MONITORING_BASE_DIR}/bin/check_disk.php > /dev/null 2>&1",
|
||||
"*/5 * * * * php {$MONITORING_BASE_DIR}/bin/check_ram.php > /dev/null 2>&1",
|
||||
"10 3 * * * php {$MONITORING_BASE_DIR}/bin/monitoring-update.php > /dev/null 2>&1",
|
||||
"* * * * * php {$MONITORING_BASE_DIR}/bin/alert-engine.php > /dev/null 2>&1"
|
||||
];
|
||||
|
||||
// Récupération du crontab actuel de root
|
||||
$current_cron = shell_exec("crontab -l 2>/dev/null") ?: "";
|
||||
$lines = explode("\n", trim($current_cron));
|
||||
$updated = false;
|
||||
|
||||
foreach ($required_jobs as $job) {
|
||||
// On extrait la commande sans les arguments de temps pour la recherche
|
||||
// On cherche si le chemin du script est déjà présent
|
||||
$script_path = explode(' ', $job)[5];
|
||||
|
||||
$found = false;
|
||||
foreach ($lines as $line) {
|
||||
if (strpos($line, $script_path) !== false) {
|
||||
$found = true;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
if (!$found) {
|
||||
log_notice("cron_added", "Ajout d'une tâche au crontab", ["job" => $job]);
|
||||
$lines[] = $job;
|
||||
$updated = true;
|
||||
}
|
||||
}
|
||||
|
||||
if ($updated) {
|
||||
// Réécriture du crontab
|
||||
$tmp_cron = tempnam(sys_get_temp_dir(), 'cron');
|
||||
file_put_contents($tmp_cron, implode("\n", $lines) . "\n");
|
||||
exec("crontab " . escapeshellarg($tmp_cron));
|
||||
unlink($tmp_cron);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* Suppression des fichiers obsolètes
|
||||
*/
|
||||
@@ -209,6 +257,7 @@ foreach ($manifest as $item) {
|
||||
|
||||
delete_extra_files($remote_paths);
|
||||
run_local_conf_sync();
|
||||
ensure_crontab_entries();
|
||||
|
||||
if ($failed > 0) {
|
||||
log_warning("update_finished_with_errors", "Mise à jour terminée avec erreurs", ["total=$total", "failed=$failed"]);
|
||||
|
||||
Reference in New Issue
Block a user