deploy-felhom-compose/controller/internal/backup/restic.go

package backup

import (
	"context"
	"crypto/rand"
	"encoding/base64"
	"encoding/json"
	"fmt"
	"log"
	"os"
	"os/exec"
	"path/filepath"
	"strings"
	"time"

	"gitea.dooplex.hu/admin/felhom-controller/internal/config"
)

// ResticManager handles restic backup operations.
// All methods accept repoPath as parameter to support per-drive repos.
type ResticManager struct {
	passwordFile string
	logger       *log.Logger
	customerID   string
	cacheDir     string
}

// SnapshotResult holds the outcome of a restic backup.
type SnapshotResult struct {
	SnapshotID   string
	FilesNew     int
	FilesChanged int
	DataAdded    string
	Duration     time.Duration
}

// SnapshotInfo holds information about a restic snapshot.
type SnapshotInfo struct {
	ID         string    `json:"short_id"`
	Time       time.Time `json:"time"`
	Paths      []string  `json:"paths"`
	Tags       []string  `json:"tags"`
	RepoPath   string    `json:"-"`            // set by caller for multi-repo aggregation
	Tier       int       `json:"tier"`         // 1 = primary, 2 = secondary
	DriveLabel string    `json:"drive_label"`  // filled by caller from settings
}

// RepoStats holds repository statistics.
type RepoStats struct {
	TotalSize      string
	TotalSizeBytes int64
	SnapshotCount  int
	LatestSnapshot *SnapshotInfo
}

// NewResticManager creates a new restic manager.
func NewResticManager(cfg *config.Config, logger *log.Logger) *ResticManager {
	return &ResticManager{
		passwordFile: cfg.Backup.ResticPasswordFile,
		logger:       logger,
		customerID:   cfg.Customer.ID,
		cacheDir:     filepath.Join(cfg.Paths.DataDir, "restic-cache"),
	}
}

// EnsureInitialized checks if the restic repo exists and initializes it if not.
// Also auto-generates the password file if missing.
func (r *ResticManager) EnsureInitialized(repoPath string) error {
	// Ensure password file exists
	if _, err := os.Stat(r.passwordFile); os.IsNotExist(err) {
		if err := r.generatePassword(); err != nil {
			return fmt.Errorf("generating restic password: %w", err)
		}
	}

	// Ensure cache dir exists
	os.MkdirAll(r.cacheDir, 0700)

	// Check if repo is already initialized
	configPath := filepath.Join(repoPath, "config")
	if _, err := os.Stat(configPath); err == nil {
		r.logger.Printf("[INFO] Restic repo already initialized at %s", repoPath)
		return nil
	}

	// Ensure repo directory exists
	if err := os.MkdirAll(repoPath, 0700); err != nil {
		return fmt.Errorf("creating repo dir: %w", err)
	}

	// Initialize repo
	r.logger.Printf("[INFO] Initializing restic repository at %s", repoPath)
	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
	defer cancel()

	cmd := r.command(ctx, repoPath, "init")
	out, err := cmd.CombinedOutput()
	if err != nil {
		return fmt.Errorf("restic init failed: %v — %s", err, truncate(string(out), 200))
	}

	r.logger.Printf("[INFO] Restic repository initialized successfully")
	return nil
}

// Snapshot creates a new backup snapshot of the given paths.
func (r *ResticManager) Snapshot(repoPath string, paths []string, tags []string) (*SnapshotResult, error) {
	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
	defer cancel()

	start := time.Now()

	args := []string{"backup", "--json"}
	for _, tag := range tags {
		args = append(args, "--tag", tag)
	}
	args = append(args, "--host", r.customerID)

	// Only include paths that exist
	var existingPaths []string
	for _, p := range paths {
		if _, err := os.Stat(p); err == nil {
			existingPaths = append(existingPaths, p)
		} else {
			r.logger.Printf("[WARN] Backup path does not exist, skipping: %s", p)
		}
	}

	if len(existingPaths) == 0 {
		return nil, fmt.Errorf("no backup paths exist")
	}
	args = append(args, existingPaths...)

	cmd := r.command(ctx, repoPath, args...)
	out, err := cmd.Output()
	if err != nil {
		// Check for stale lock
		errStr := string(out)
		if strings.Contains(errStr, "lock") || strings.Contains(errStr, "locked") {
			r.logger.Printf("[WARN] Restic repo locked — attempting unlock")
			unlockCmd := r.command(ctx, repoPath, "unlock")
			unlockCmd.Run()
			// Retry once
			cmd = r.command(ctx, repoPath, args...)
			out, err = cmd.Output()
			if err != nil {
				return nil, fmt.Errorf("restic backup failed after unlock: %v", err)
			}
		} else {
			return nil, fmt.Errorf("restic backup failed: %v", err)
		}
	}

	result := &SnapshotResult{
		Duration: time.Since(start),
	}

	// Parse JSON output — look for the summary line
	for _, line := range strings.Split(string(out), "\n") {
		line = strings.TrimSpace(line)
		if line == "" {
			continue
		}

		var msg struct {
			MessageType  string `json:"message_type"`
			FilesNew     int    `json:"files_new"`
			FilesChanged int    `json:"files_changed"`
			DataAdded    int64  `json:"data_added"`
			SnapshotID   string `json:"snapshot_id"`
		}
		if err := json.Unmarshal([]byte(line), &msg); err != nil {
			continue
		}
		if msg.MessageType == "summary" {
			result.SnapshotID = msg.SnapshotID
			result.FilesNew = msg.FilesNew
			result.FilesChanged = msg.FilesChanged
			result.DataAdded = humanizeBytes(msg.DataAdded)
		}
	}

	return result, nil
}

// Prune removes old snapshots according to retention policy.
func (r *ResticManager) Prune(repoPath string, retention config.RetentionConfig) error {
	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
	defer cancel()

	args := []string{
		"forget",
		"--keep-daily", fmt.Sprintf("%d", retention.KeepDaily),
		"--keep-weekly", fmt.Sprintf("%d", retention.KeepWeekly),
		"--keep-monthly", fmt.Sprintf("%d", retention.KeepMonthly),
		"--prune",
	}

	cmd := r.command(ctx, repoPath, args...)
	out, err := cmd.CombinedOutput()
	if err != nil {
		return fmt.Errorf("restic forget/prune failed: %v — %s", err, truncate(string(out), 200))
	}

	r.logger.Printf("[INFO] Restic prune completed for %s", repoPath)
	return nil
}

// Check verifies repository integrity.
func (r *ResticManager) Check(repoPath string) error {
	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
	defer cancel()

	cmd := r.command(ctx, repoPath, "check")
	out, err := cmd.CombinedOutput()
	if err != nil {
		return fmt.Errorf("restic check failed: %v — %s", err, truncate(string(out), 200))
	}
	return nil
}

// ListSnapshots returns all snapshots, newest first, limited to N entries.
func (r *ResticManager) ListSnapshots(repoPath string, limit int) ([]SnapshotInfo, error) {
	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
	defer cancel()

	cmd := r.command(ctx, repoPath, "snapshots", "--json")
	out, err := cmd.Output()
	if err != nil {
		return nil, fmt.Errorf("restic snapshots failed: %v", err)
	}

	var snapshots []SnapshotInfo
	if err := json.Unmarshal(out, &snapshots); err != nil {
		return nil, fmt.Errorf("parsing snapshot JSON: %v", err)
	}

	// Reverse for newest first
	for i, j := 0, len(snapshots)-1; i < j; i, j = i+1, j-1 {
		snapshots[i], snapshots[j] = snapshots[j], snapshots[i]
	}

	if limit > 0 && len(snapshots) > limit {
		snapshots = snapshots[:limit]
	}

	return snapshots, nil
}

// LatestSnapshot returns the most recent snapshot info.
func (r *ResticManager) LatestSnapshot(repoPath string) (*SnapshotInfo, error) {
	ctx, cancel := context.WithTimeout(context.Background(), 1*time.Minute)
	defer cancel()

	cmd := r.command(ctx, repoPath, "snapshots", "--latest", "1", "--json")
	out, err := cmd.Output()
	if err != nil {
		return nil, fmt.Errorf("restic snapshots failed: %v", err)
	}

	var snapshots []SnapshotInfo
	if err := json.Unmarshal(out, &snapshots); err != nil {
		return nil, fmt.Errorf("parsing snapshot JSON: %v", err)
	}

	if len(snapshots) == 0 {
		return nil, nil
	}

	return &snapshots[0], nil
}

// Stats returns repository statistics.
func (r *ResticManager) Stats(repoPath string) (*RepoStats, error) {
	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
	defer cancel()

	stats := &RepoStats{}

	// Get repo size
	cmd := r.command(ctx, repoPath, "stats", "--json")
	out, err := cmd.Output()
	if err == nil {
		var raw struct {
			TotalSize uint64 `json:"total_size"`
		}
		if json.Unmarshal(out, &raw) == nil {
			stats.TotalSizeBytes = int64(raw.TotalSize)
			stats.TotalSize = humanizeBytes(stats.TotalSizeBytes)
		}
	}

	// Count snapshots
	cmd = r.command(ctx, repoPath, "snapshots", "--json")
	out, err = cmd.Output()
	if err == nil {
		var snapshots []SnapshotInfo
		if json.Unmarshal(out, &snapshots) == nil {
			stats.SnapshotCount = len(snapshots)
			if len(snapshots) > 0 {
				latest := snapshots[len(snapshots)-1]
				stats.LatestSnapshot = &latest
			}
		}
	}

	return stats, nil
}

// GetPassword reads and returns the restic repository password.
func (r *ResticManager) GetPassword() (string, error) {
	data, err := os.ReadFile(r.passwordFile)
	if err != nil {
		return "", fmt.Errorf("reading restic password: %w", err)
	}
	return strings.TrimSpace(string(data)), nil
}

// RestoreAppData restores specific paths from a restic snapshot.
func (r *ResticManager) RestoreAppData(repoPath string, snapshotID string, paths []string) error {
	ctx, cancel := context.WithTimeout(context.Background(), 30*time.Minute)
	defer cancel()

	args := []string{
		"restore", snapshotID,
		"--target", "/",
	}
	for _, p := range paths {
		args = append(args, "--include", p)
	}

	r.logger.Printf("[WARN] RESTORE started: repo=%s, snapshot=%s, paths=%v", repoPath, snapshotID, paths)

	cmd := r.command(ctx, repoPath, args...)
	output, err := cmd.CombinedOutput()
	if err != nil {
		r.logger.Printf("[ERROR] Restore failed: %v, output: %s", err, truncate(string(output), 500))
		return fmt.Errorf("restic restore failed: %w", err)
	}

	r.logger.Printf("[INFO] RESTORE completed: snapshot=%s, paths=%v", snapshotID, paths)
	return nil
}

// RepoExists checks if a restic repo is initialized at the given path.
func (r *ResticManager) RepoExists(repoPath string) bool {
	_, err := os.Stat(filepath.Join(repoPath, "config"))
	return err == nil
}

// UnlockCommand returns an exec.Cmd that runs restic unlock on the given repo.
func (r *ResticManager) UnlockCommand(ctx context.Context, repoPath string) *exec.Cmd {
	return r.command(ctx, repoPath, "unlock")
}

func (r *ResticManager) command(ctx context.Context, repoPath string, args ...string) *exec.Cmd {
	cmd := exec.CommandContext(ctx, "restic", args...)
	cmd.Env = append(os.Environ(),
		"RESTIC_REPOSITORY="+repoPath,
		"RESTIC_PASSWORD_FILE="+r.passwordFile,
		"RESTIC_CACHE_DIR="+r.cacheDir,
	)
	return cmd
}

func (r *ResticManager) generatePassword() error {
	// Ensure directory exists
	dir := filepath.Dir(r.passwordFile)
	if err := os.MkdirAll(dir, 0700); err != nil {
		return fmt.Errorf("creating password dir: %w", err)
	}

	// Generate 32 random bytes, base64url-encode
	b := make([]byte, 32)
	if _, err := rand.Read(b); err != nil {
		return fmt.Errorf("generating random bytes: %w", err)
	}
	password := base64.URLEncoding.EncodeToString(b)

	if err := os.WriteFile(r.passwordFile, []byte(password), 0600); err != nil {
		return fmt.Errorf("writing password file: %w", err)
	}

	r.logger.Printf("[INFO] Generated new restic repository password at %s", r.passwordFile)
	r.logger.Printf("[WARN] Save this password externally — losing it means losing access to ALL backups")
	return nil
}

func truncate(s string, maxLen int) string {
	if len(s) <= maxLen {
		return s
	}
	return s[:maxLen] + "..."
}