canon/internal/backup/backup.go

// Package backup writes the database and cached images to .tar.gz archives,
// on demand and on a cron schedule.
package backup

import (
	"archive/tar"
	"compress/gzip"
	"context"
	"errors"
	"fmt"
	"io"
	"log/slog"
	"os"
	"path/filepath"
	"regexp"
	"slices"
	"strconv"
	"strings"
	"sync"
	"time"

	"github.com/cdnutter/canon/internal/store"
)

const (
	DefaultCron = "0 3 * * *" // every day at 3:00
	DefaultKeep = 7
	MaxKeep     = 365
)

var ErrNotFound = errors.New("backup not found")

// Scheduled backups are named canon-YYYYMMDD-HHMMSS.tar.gz; ones made with
// "Back up now" end in -manual and are never pruned automatically.
var nameRe = regexp.MustCompile(`^canon-\d{8}-\d{6}(-\d+)?(-manual)?\.tar\.gz$`)

type Info struct {
	Name    string    `json:"name"`
	Size    int64     `json:"size"`
	Created time.Time `json:"created"`
	Manual  bool      `json:"manual"`
}

type Status struct {
	LastRun time.Time `json:"lastRun,omitzero"`
	LastErr string    `json:"lastError,omitempty"`
}

type Manager struct {
	dir       string
	imagesDir string
	store     *store.Store
	log       *slog.Logger

	run    sync.Mutex // one backup at a time
	reload chan struct{}

	mu     sync.Mutex
	status Status
}

func New(dir, imagesDir string, st *store.Store, log *slog.Logger) *Manager {
	return &Manager{dir: dir, imagesDir: imagesDir, store: st, log: log, reload: make(chan struct{}, 1)}
}

func (m *Manager) Dir() string { return m.dir }

func (m *Manager) Status() Status {
	m.mu.Lock()
	defer m.mu.Unlock()
	return m.status
}

// Reschedule makes the scheduler re-read its settings.
func (m *Manager) Reschedule() {
	select {
	case m.reload <- struct{}{}:
	default:
	}
}

// Settings returns the configured schedule and retention, with defaults.
func Settings(set map[string]string) (cron string, keep int) {
	cron = strings.TrimSpace(set[store.SettingBackupCron])
	if cron == "" {
		cron = DefaultCron
	}
	keep = DefaultKeep
	if n, err := strconv.Atoi(set[store.SettingBackupKeep]); err == nil && n > 0 {
		keep = n
	}
	return cron, keep
}

// NextRun is when the schedule next fires after t, or nil when backups are
// off or the schedule is invalid.
func NextRun(cron string, t time.Time) *time.Time {
	if cron == "off" {
		return nil
	}
	sched, err := ParseCron(cron)
	if err != nil {
		return nil
	}
	next := sched.Next(t)
	if next.IsZero() {
		return nil
	}
	return &next
}

// Run backs up on schedule until ctx is canceled.
func (m *Manager) Run(ctx context.Context) {
	for {
		set, err := m.store.Settings(ctx)
		if err != nil {
			m.log.Warn("backup: loading settings", "err", err)
		}
		cron, keep := Settings(set)
		var next time.Time
		if n := NextRun(cron, time.Now()); n != nil {
			next = *n
		} else if cron != "off" {
			m.log.Warn("backup: invalid schedule, not backing up", "cron", cron)
		}

		// Wake at least hourly so a changed clock or a suspended host
		// doesn't delay a backup by long.
		wait := time.Hour
		if !next.IsZero() {
			wait = min(wait, time.Until(next))
		}
		timer := time.NewTimer(max(wait, 0))
		select {
		case <-ctx.Done():
			timer.Stop()
			return
		case <-m.reload:
			timer.Stop()
			continue
		case <-timer.C:
		}
		if next.IsZero() || time.Now().Before(next) {
			continue
		}
		if _, err := m.Create(ctx, false); err != nil {
			m.log.Error("scheduled backup failed", "err", err)
			continue
		}
		if err := m.prune(keep); err != nil {
			m.log.Warn("backup: removing old backups", "err", err)
		}
	}
}

// Create writes a new backup and returns it.
func (m *Manager) Create(ctx context.Context, manual bool) (info Info, err error) {
	m.run.Lock()
	defer m.run.Unlock()
	defer func() {
		m.mu.Lock()
		m.status.LastRun = time.Now()
		m.status.LastErr = ""
		if err != nil {
			m.status.LastErr = err.Error()
		}
		m.mu.Unlock()
	}()

	if err := os.MkdirAll(m.dir, 0o755); err != nil {
		return info, err
	}
	name := m.freeName(time.Now(), manual)
	tmpDir, err := os.MkdirTemp(m.dir, ".tmp-")
	if err != nil {
		return info, err
	}
	defer os.RemoveAll(tmpDir)

	dbCopy := filepath.Join(tmpDir, "canon.db")
	if err := m.store.SnapshotTo(ctx, dbCopy); err != nil {
		return info, fmt.Errorf("copying database: %w", err)
	}
	partial := filepath.Join(tmpDir, name)
	if err := m.writeArchive(partial, dbCopy); err != nil {
		return info, err
	}
	final := filepath.Join(m.dir, name)
	if err := os.Rename(partial, final); err != nil {
		return info, err
	}
	fi, err := os.Stat(final)
	if err != nil {
		return info, err
	}
	info = Info{Name: name, Size: fi.Size(), Created: fi.ModTime(), Manual: manual}
	m.log.Info("backup written", "file", final, "size", fi.Size(), "manual", manual)
	return info, nil
}

func (m *Manager) freeName(t time.Time, manual bool) string {
	base := "canon-" + t.Format("20060102-150405")
	suffix := ".tar.gz"
	if manual {
		suffix = "-manual.tar.gz"
	}
	name := base + suffix
	for i := 2; ; i++ {
		if _, err := os.Stat(filepath.Join(m.dir, name)); errors.Is(err, os.ErrNotExist) {
			return name
		}
		name = fmt.Sprintf("%s-%d%s", base, i, suffix)
	}
}

func (m *Manager) writeArchive(path, dbCopy string) (err error) {
	f, err := os.Create(path)
	if err != nil {
		return err
	}
	defer func() {
		if cerr := f.Close(); err == nil {
			err = cerr
		}
	}()
	gz := gzip.NewWriter(f)
	tw := tar.NewWriter(gz)
	if err := addFile(tw, dbCopy, "canon.db"); err != nil {
		return err
	}
	entries, err := os.ReadDir(m.imagesDir)
	if err != nil && !errors.Is(err, os.ErrNotExist) {
		return err
	}
	for _, e := range entries {
		if !e.Type().IsRegular() || strings.HasPrefix(e.Name(), ".") {
			continue // skip in-progress uploads
		}
		if err := addFile(tw, filepath.Join(m.imagesDir, e.Name()), "images/"+e.Name()); err != nil {
			if errors.Is(err, os.ErrNotExist) {
				continue // removed while we were working
			}
			return err
		}
	}
	if err := tw.Close(); err != nil {
		return err
	}
	if err := gz.Close(); err != nil {
		return err
	}
	return f.Sync()
}

func addFile(tw *tar.Writer, src, name string) error {
	f, err := os.Open(src)
	if err != nil {
		return err
	}
	defer f.Close()
	fi, err := f.Stat()
	if err != nil {
		return err
	}
	hdr := &tar.Header{Name: name, Mode: 0o644, Size: fi.Size(), ModTime: fi.ModTime(), Typeflag: tar.TypeReg}
	if err := tw.WriteHeader(hdr); err != nil {
		return err
	}
	_, err = io.CopyN(tw, f, fi.Size())
	return err
}

// List returns backups, newest first.
func (m *Manager) List() ([]Info, error) {
	entries, err := os.ReadDir(m.dir)
	if errors.Is(err, os.ErrNotExist) {
		return []Info{}, nil
	}
	if err != nil {
		return nil, err
	}
	out := []Info{}
	for _, e := range entries {
		if !e.Type().IsRegular() || !nameRe.MatchString(e.Name()) {
			continue
		}
		fi, err := e.Info()
		if err != nil {
			continue
		}
		out = append(out, Info{Name: e.Name(), Size: fi.Size(), Created: fi.ModTime(), Manual: strings.HasSuffix(e.Name(), "-manual.tar.gz")})
	}
	slices.SortFunc(out, func(a, b Info) int { return b.Created.Compare(a.Created) })
	return out, nil
}

// Path returns the file for a backup name, refusing anything that isn't one.
func (m *Manager) Path(name string) (string, error) {
	if !nameRe.MatchString(name) {
		return "", ErrNotFound
	}
	p := filepath.Join(m.dir, name)
	if _, err := os.Stat(p); err != nil {
		return "", ErrNotFound
	}
	return p, nil
}

func (m *Manager) Delete(name string) error {
	p, err := m.Path(name)
	if err != nil {
		return err
	}
	return os.Remove(p)
}

// prune keeps the newest keep scheduled backups. Manual ones stay until
// deleted by hand.
func (m *Manager) prune(keep int) error {
	all, err := m.List()
	if err != nil {
		return err
	}
	n := 0
	for _, b := range all {
		if b.Manual {
			continue
		}
		if n++; n > keep {
			if err := os.Remove(filepath.Join(m.dir, b.Name)); err != nil {
				return err
			}
			m.log.Info("removed old backup", "file", b.Name)
		}
	}
	return nil
}