#!/bin/bash
################################################################################
# Script Name: caprover-exporter.sh
# Version: 1.0
# Description: Prometheus exporter for CapRover PaaS providing operational
#              metrics via the CapRover API — app deployment status, container
#              health, resource usage, and platform metrics
#
# Author: Phil Connor
# Contact: contact@mylinux.work
# Website: https://mylinux.work
# License: MIT
#
# Prerequisites:
#   - CapRover installed and running
#   - CapRover API accessible (default: http://localhost:3000)
#   - curl for API calls
#   - jq for JSON parsing
#   - netcat (nc) for HTTP mode
#
# Usage:
#   ./caprover-exporter.sh                         # Output to stdout
#   ./caprover-exporter.sh --http -p 9196          # HTTP server mode
#   ./caprover-exporter.sh --textfile              # Textfile collector mode
#   ./caprover-exporter.sh --password secret       # Custom password
#
# Metrics Exported:
#   - caprover_up - API reachability (1=up, 0=down)
#   - caprover_info{version} - CapRover version info
#   - caprover_apps_total - Total app count
#   - caprover_apps_running - Running app count
#   - caprover_apps_stopped - Stopped app count
#   - caprover_app_running{app} - Per-app running status (1/0)
#   - caprover_app_instance_count{app} - Per-app replica count
#   - caprover_app_has_ssl{app} - Per-app SSL status (1/0)
#   - caprover_app_force_ssl{app} - Per-app force SSL status (1/0)
#   - caprover_nodes_total - Swarm node count
#   - caprover_volumes_total - Docker volume count
#   - caprover_disk_used_bytes - Disk usage in bytes
#   - caprover_disk_total_bytes - Total disk in bytes
#   - caprover_exporter_duration_seconds - Script execution time
#   - caprover_exporter_last_run_timestamp - Last run timestamp
#
# Configuration:
#   Default HTTP port: 9196
#   Default CapRover URL: http://localhost:3000
#   Default password: captain42
#   Textfile directory: /var/lib/node_exporter
#
################################################################################

set -euo pipefail

# ============================================================================
# CONFIGURATION VARIABLES
# ============================================================================

TEXTFILE_DIR="/var/lib/node_exporter"
OUTPUT_FILE=""
HTTP_MODE=false
HTTP_PORT=9196
CAPROVER_URL="http://localhost:3000"
CAPROVER_PASSWORD="captain42"
AUTH_TOKEN=""

# ============================================================================
# HELPER FUNCTIONS
# ============================================================================

show_usage() {
    cat <<EOF
Usage: $0 [OPTIONS]

Export CapRover PaaS statistics as Prometheus metrics via the CapRover API.

MODES:
    --textfile      Write to node_exporter textfile collector
    --http          Run HTTP server on port $HTTP_PORT

OPTIONS:
    -p, --port           HTTP port (default: 9196)
    -o, --output         Output file path
    --caprover-url       CapRover API URL (default: http://localhost:3000)
    --password           CapRover password (default: captain42)

EXAMPLES:
    $0 --textfile                                    # Write to textfile collector
    $0 --http --port 9196                            # Run HTTP server
    $0 --caprover-url https://captain.example.com    # Custom URL
    $0 --password mysecret                           # Custom password

EOF
    exit 0
}

parse_args() {
    while [[ $# -gt 0 ]]; do
        case $1 in
            -h|--help) show_usage ;;
            --textfile) OUTPUT_FILE="$TEXTFILE_DIR/caprover.prom"; shift ;;
            --http) HTTP_MODE=true; shift ;;
            -p|--port) HTTP_PORT="$2"; shift 2 ;;
            -o|--output) OUTPUT_FILE="$2"; shift 2 ;;
            --caprover-url) CAPROVER_URL="$2"; shift 2 ;;
            --password) CAPROVER_PASSWORD="$2"; shift 2 ;;
            *) echo "Unknown option: $1" >&2; exit 1 ;;
        esac
    done
}

check_prerequisites() {
    if ! command -v curl >/dev/null 2>&1; then
        echo "ERROR: curl not found" >&2; return 1
    fi
    if ! command -v jq >/dev/null 2>&1; then
        echo "ERROR: jq not found (required for JSON parsing)" >&2; return 1
    fi
    return 0
}

prom_escape() {
    local val="$1"
    val="${val//\\/\\\\}"
    val="${val//\"/\\\"}"
    val="${val//$'\n'/}"
    echo "$val"
}

authenticate() {
    if [ -n "$AUTH_TOKEN" ]; then return 0; fi

    local response
    response=$(curl -s -X POST \
        -H "Content-Type: application/json" \
        -H "x-namespace: captain" \
        -d "{\"password\":\"${CAPROVER_PASSWORD}\"}" \
        "${CAPROVER_URL}/api/v2/login" 2>/dev/null) || return 1

    local status
    status=$(echo "$response" | jq -r '.status // 0' 2>/dev/null)
    if [ "$status" != "100" ]; then
        echo "ERROR: Failed to authenticate with CapRover API" >&2; return 1
    fi

    AUTH_TOKEN=$(echo "$response" | jq -r '.data.token // empty' 2>/dev/null)
    if [ -z "$AUTH_TOKEN" ]; then
        echo "ERROR: No auth token received from CapRover API" >&2; return 1
    fi
    return 0
}

api_call() {
    local endpoint="$1"
    curl -s -X POST \
        -H "Content-Type: application/json" \
        -H "x-namespace: captain" \
        -H "x-captain-auth: ${AUTH_TOKEN}" \
        -d "{}" \
        "${CAPROVER_URL}${endpoint}" 2>/dev/null
}

api_get() {
    local endpoint="$1"
    curl -s -X GET \
        -H "Content-Type: application/json" \
        -H "x-namespace: captain" \
        -H "x-captain-auth: ${AUTH_TOKEN}" \
        "${CAPROVER_URL}${endpoint}" 2>/dev/null
}

# ============================================================================
# METRIC GENERATION
# ============================================================================

generate_metrics() {
    local script_start
    script_start=$(date +%s)

    if ! check_prerequisites; then
        echo "# HELP caprover_up CapRover API reachability (1=up, 0=down)"
        echo "# TYPE caprover_up gauge"
        echo "caprover_up 0"
        return
    fi

    AUTH_TOKEN=""
    if ! authenticate; then
        echo "# HELP caprover_up CapRover API reachability (1=up, 0=down)"
        echo "# TYPE caprover_up gauge"
        echo "caprover_up 0"
        return
    fi

    cat <<EOF
# HELP caprover_up CapRover API reachability (1=up, 0=down)
# TYPE caprover_up gauge
caprover_up 1

EOF

    # ========================================================================
    # SYSTEM INFO
    # ========================================================================

    local system_info
    system_info=$(api_get "/api/v2/user/system/info")

    local caprover_version="unknown" node_count=0 disk_used=0 disk_total=0 volume_count=0

    if [ -n "$system_info" ] && [ "$system_info" != "null" ]; then
        local info_status
        info_status=$(echo "$system_info" | jq -r '.status // 0' 2>/dev/null)
        if [ "$info_status" = "100" ]; then
            caprover_version=$(echo "$system_info" | jq -r '.data.caproverVersion // "unknown"' 2>/dev/null)
            node_count=$(echo "$system_info" | jq -r '.data.swarmNodesCount // 0' 2>/dev/null)
            node_count=${node_count:-0}
            disk_used=$(echo "$system_info" | jq -r '.data.diskUsedInMb // 0' 2>/dev/null)
            disk_total=$(echo "$system_info" | jq -r '.data.diskTotalInMb // 0' 2>/dev/null)
            disk_used=${disk_used:-0}; disk_total=${disk_total:-0}
            [ "$disk_used" != "0" ] && disk_used=$((disk_used * 1024 * 1024))
            [ "$disk_total" != "0" ] && disk_total=$((disk_total * 1024 * 1024))
            volume_count=$(echo "$system_info" | jq -r '.data.dockerVolumesCount // 0' 2>/dev/null)
            volume_count=${volume_count:-0}
        fi
    fi

    cat <<EOF
# HELP caprover_info CapRover version information
# TYPE caprover_info gauge
caprover_info{version="$(prom_escape "$caprover_version")"} 1

# HELP caprover_nodes_total Number of swarm nodes
# TYPE caprover_nodes_total gauge
caprover_nodes_total $node_count

# HELP caprover_volumes_total Number of Docker volumes
# TYPE caprover_volumes_total gauge
caprover_volumes_total $volume_count

# HELP caprover_disk_used_bytes Disk usage in bytes
# TYPE caprover_disk_used_bytes gauge
caprover_disk_used_bytes $disk_used

# HELP caprover_disk_total_bytes Total disk in bytes
# TYPE caprover_disk_total_bytes gauge
caprover_disk_total_bytes $disk_total

EOF

    # ========================================================================
    # APP METRICS
    # ========================================================================

    local apps_response
    apps_response=$(api_call "/api/v2/user/apps/appDefinitions")

    local total_apps=0 running_apps=0 stopped_apps=0

    if [ -n "$apps_response" ] && [ "$apps_response" != "null" ]; then
        local apps_status
        apps_status=$(echo "$apps_response" | jq -r '.status // 0' 2>/dev/null)

        if [ "$apps_status" = "100" ]; then
            total_apps=$(echo "$apps_response" | jq '.data.appDefinitions | length // 0' 2>/dev/null)
            total_apps=${total_apps:-0}
            running_apps=$(echo "$apps_response" | jq '[.data.appDefinitions[] | select(.deployedVersion != null and .deployedVersion != 0)] | length' 2>/dev/null)
            stopped_apps=$(echo "$apps_response" | jq '[.data.appDefinitions[] | select(.deployedVersion == null or .deployedVersion == 0)] | length' 2>/dev/null)
            running_apps=${running_apps:-0}; stopped_apps=${stopped_apps:-0}

            cat <<EOF
# HELP caprover_apps_total Total number of deployed apps
# TYPE caprover_apps_total gauge
caprover_apps_total $total_apps

# HELP caprover_apps_running Apps with running containers
# TYPE caprover_apps_running gauge
caprover_apps_running $running_apps

# HELP caprover_apps_stopped Apps with stopped containers
# TYPE caprover_apps_stopped gauge
caprover_apps_stopped $stopped_apps

EOF

            # Per-app running status
            echo "# HELP caprover_app_running Per-app running status (1=running, 0=stopped)"
            echo "# TYPE caprover_app_running gauge"
            echo "$apps_response" | jq -r '.data.appDefinitions[] | "\(.appName) \(if .deployedVersion != null and .deployedVersion != 0 then 1 else 0 end)"' 2>/dev/null | while read -r name val; do
                [ -z "$name" ] && continue
                echo "caprover_app_running{app=\"$(prom_escape "$name")\"} $val"
            done
            echo ""

            # Per-app instance count
            echo "# HELP caprover_app_instance_count Number of replicas per app"
            echo "# TYPE caprover_app_instance_count gauge"
            echo "$apps_response" | jq -r '.data.appDefinitions[] | "\(.appName) \(.instanceCount // 1)"' 2>/dev/null | while read -r name val; do
                [ -z "$name" ] && continue
                echo "caprover_app_instance_count{app=\"$(prom_escape "$name")\"} $val"
            done
            echo ""

            # Per-app SSL status
            echo "# HELP caprover_app_has_ssl SSL enabled per app (1=yes, 0=no)"
            echo "# TYPE caprover_app_has_ssl gauge"
            echo "$apps_response" | jq -r '.data.appDefinitions[] | "\(.appName) \(if .hasDefaultSubDomainSsl == true then 1 else 0 end)"' 2>/dev/null | while read -r name val; do
                [ -z "$name" ] && continue
                echo "caprover_app_has_ssl{app=\"$(prom_escape "$name")\"} $val"
            done
            echo ""

            # Per-app force SSL status
            echo "# HELP caprover_app_force_ssl Force SSL per app (1=yes, 0=no)"
            echo "# TYPE caprover_app_force_ssl gauge"
            echo "$apps_response" | jq -r '.data.appDefinitions[] | "\(.appName) \(if .forceSsl == true then 1 else 0 end)"' 2>/dev/null | while read -r name val; do
                [ -z "$name" ] && continue
                echo "caprover_app_force_ssl{app=\"$(prom_escape "$name")\"} $val"
            done
        else
            echo "# HELP caprover_apps_total Total number of deployed apps"
            echo "# TYPE caprover_apps_total gauge"
            echo "caprover_apps_total 0"
        fi
    else
        echo "# HELP caprover_apps_total Total number of deployed apps"
        echo "# TYPE caprover_apps_total gauge"
        echo "caprover_apps_total 0"
    fi

    echo ""

    # ========================================================================
    # EXPORTER RUNTIME
    # ========================================================================

    local script_end script_duration
    script_end=$(date +%s)
    script_duration=$((script_end - script_start))

    cat <<EOF
# HELP caprover_exporter_duration_seconds Time to generate all metrics
# TYPE caprover_exporter_duration_seconds gauge
caprover_exporter_duration_seconds $script_duration

# HELP caprover_exporter_last_run_timestamp Unix timestamp of last successful run
# TYPE caprover_exporter_last_run_timestamp gauge
caprover_exporter_last_run_timestamp $script_end
EOF
}

# ============================================================================
# HTTP SERVER MODE
# ============================================================================

run_http_server() {
    echo "Starting CapRover exporter on port $HTTP_PORT..." >&2

    if ! command -v nc >/dev/null 2>&1; then
        echo "ERROR: netcat (nc) required for HTTP mode" >&2
        exit 1
    fi

    while true; do
        {
            read -r request
            if [[ "$request" =~ ^GET\ /metrics ]]; then
                echo -e "HTTP/1.1 200 OK\r\nContent-Type: text/plain; version=0.0.4\r\n\r"
                generate_metrics
            else
                echo -e "HTTP/1.1 200 OK\r\nContent-Type: text/html\r\n\r"
                cat <<EOF
<!DOCTYPE html>
<html>
<head><title>CapRover Exporter v1.0</title></head>
<body>
<h1>CapRover Prometheus Exporter v1.0</h1>
<p><a href="/metrics">Metrics</a></p>
<p>Operational metrics from the CapRover API.</p>
</body>
</html>
EOF
            fi
        } | nc -l -p "$HTTP_PORT" -q 1 2>/dev/null
    done
}

# ============================================================================
# MAIN EXECUTION
# ============================================================================

main() {
    parse_args "$@"

    if [ "$HTTP_MODE" = true ]; then
        run_http_server
    elif [ -n "$OUTPUT_FILE" ]; then
        local output_dir
        output_dir="$(dirname "$OUTPUT_FILE")"
        mkdir -p "$output_dir"

        local temp_file
        temp_file=$(mktemp "${output_dir}/.caprover_metrics.XXXXXX")

        if ! generate_metrics > "$temp_file" 2>/dev/null; then
            rm -f "$temp_file"
            echo "ERROR: Failed to generate metrics" >&2
            exit 1
        fi

        local file_lines
        file_lines=$(wc -l < "$temp_file" 2>/dev/null || echo 0)

        if [ "$file_lines" -lt 10 ]; then
            rm -f "$temp_file"
            echo "ERROR: Metrics file too small ($file_lines lines), keeping previous" >&2
            exit 1
        fi

        chmod 644 "$temp_file"
        mv -f "$temp_file" "$OUTPUT_FILE"
        echo "Metrics written to $OUTPUT_FILE ($file_lines lines)" >&2
    else
        generate_metrics
    fi
}

main "$@"