How to See Your SSD/HDD Health Directly in Home Assistant

#!/bin/bash

# --- CONFIGURATION ---
# Output file location on the host (where Home Assistant is mounted to /config)
OUTPUT="/opt/HA/homeassistant/smart_status.json"
DISKS="/dev/sda /dev/sdb"

# Create a temporary file for atomic write
TMP=$(mktemp)
echo "{" > "$TMP"
first=1

for disk in $DISKS; do
    # Prepend comma if not the first disk
    [ "$first" -eq 0 ] && echo "," >> "$TMP"
    first=0

    key=$(basename "$disk" | sed 's/n1$//')

    # --- 1. SET GUARANTEED DEFAULT JSON VALUES ---
    # These are set for every disk. If smartctl fails, these valid JSON primitives are used.
    model="Unknown"
    serial="Unknown"
    health="false"  # Boolean written as string literal
    temp="null"     # JSON null literal
    poh="null"      # JSON null literal
    bad="0"         # JSON number literal
    pending="0"     # JSON number literal
    wear="null"     # JSON null literal
    
    # --- Check if disk is even present ---
    if [ ! -b "$disk" ]; then
        cat <<EOT >> "$TMP"
  "$key": {
    "last_updated": "$(date -Iminutes)",
    "health": false,
    "model": "Not present",
    "serial": "",
    "temperature": null,
    "power_on_hours": null,
    "bad_blocks": 0,
    "pending_sectors": 0,
    "wear_level": null
  }
EOT
        continue
    fi

    # --- 2. TRY TO GET SMART DATA ---
    data=""
    for type in "" "-d nvme" "-d ata" "-d sat" "-d usbjmicron" "-d usbcypress" "-d sat,auto"; do
        # 12-second timeout per attempt
        data=$(timeout 12 /usr/sbin/smartctl --json=c -a $type "$disk" 2>/dev/null || true)
        # If we got a model name, the command succeeded.
        if echo "$data" | /usr/bin/jq -e '.model_name' >/dev/null 2>&1; then
            break
        fi
        data="" # Clear if check fails
    done

    # --- 3. PROCESS DATA IF AVAILABLE ---
    if [ -n "$data" ]; then
        # Overwrite defaults with actual values
        model=$(echo "$data" | /usr/bin/jq -r '.model_name // "Unknown"')
        serial=$(echo "$data" | /usr/bin/jq -r '.serial_number // "Unknown"')
        health=$(echo "$data" | /usr/bin/jq -r '.smart_status.passed // false')

        temp=$(echo "$data" | /usr/bin/jq -r '
            .temperature.current // 
            .temperature_celsius.current // 
            .nvme_smart_health_information_log.temperature // 
            null
        ')

        poh=$(echo "$data" | /usr/bin/jq -r '
            .power_on_time.hours // 
            .nvme_smart_health_information_log.power_on_hours // 
            null
        ')

        bad=$(echo "$data" | /usr/bin/jq -r '
            (.ata_smart_attributes.table // []) | 
            map(select(.name == "Reallocated_Sector_Ct")) | 
            .[0].raw.value // 0
        ')

        pending=$(echo "$data" | /usr/bin/jq -r '
            (.ata_smart_attributes.table // []) | 
            map(select(.name == "Current_Pending_Sector")) | 
            .[0].raw.value // 0
        ')

        wear=$(echo "$data" | /usr/bin/jq -r '.nvme_smart_health_information_log.percentage_used // null')
    fi

    # --- 4. WRITE FINAL JSON BLOCK ---
    cat <<EOT >> "$TMP"
  "$key": {
    "last_updated": "$(date -Iminutes)",
    "health": $health,
    "model": $(printf '%s' "$model" | jq -R .),
    "serial": $(printf '%s' "$serial" | jq -R .),
    "temperature": $temp,
    "power_on_hours": $poh,
    "bad_blocks": $bad,
    "pending_sectors": $pending,
    "wear_level": $wear
  }
EOT
done

# --- ATOMIC RENAME AND CLEANUP ---

# Close the JSON object
echo -e "\n}" >> "$TMP"

# Ensure the output directory exists
mkdir -p "$(dirname "$OUTPUT")"

# Atomically replace the old file with the new temporary file
mv "$TMP" "$OUTPUT"

# Set permissions for easy reading by the HA container
chmod 644 "$OUTPUT"

healthy=$(jq 'to_entries | map(select(.value.health == true)) | length' "$OUTPUT")
total=$(jq 'keys | length' "$OUTPUT")
echo "smart_status.json updated — $healthy/$total disks healthy → $OUTPUT"