WIP: backup revolution [ci skip]

See coop-cloud/organising#485

cli/app/backup.go

@@ -1,414 +1,425 @@
 package app
 
 import (
-	"archive/tar"
 	"context"
 	"fmt"
-	"io"
-	"os"
-	"path/filepath"
-	"strings"
-	"time"
 
 	"coopcloud.tech/abra/cli/internal"
 	"coopcloud.tech/abra/pkg/autocomplete"
 	"coopcloud.tech/abra/pkg/client"
 	"coopcloud.tech/abra/pkg/config"
 	containerPkg "coopcloud.tech/abra/pkg/container"
-	recipePkg "coopcloud.tech/abra/pkg/recipe"
+	"coopcloud.tech/abra/pkg/recipe"
+	"coopcloud.tech/abra/pkg/service"
 	"coopcloud.tech/abra/pkg/upstream/container"
 	"github.com/docker/cli/cli/command"
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/filters"
-	dockerClient "github.com/docker/docker/client"
-	"github.com/docker/docker/pkg/archive"
-	"github.com/klauspost/pgzip"
 	"github.com/sirupsen/logrus"
 	"github.com/urfave/cli"
 )
 
-type backupConfig struct {
+var snapshot string
-	preHookCmd  string
+var snapshotFlag = &cli.StringFlag{
-	postHookCmd string
+	Name:        "snapshot, s",
-	backupPaths []string
+	Usage:       "Lists specific snapshot",
+	Destination: &snapshot,
 }
 
-var appBackupCommand = cli.Command{
+var includePath string
-	Name:      "backup",
+var includePathFlag = &cli.StringFlag{
-	Aliases:   []string{"bk"},
+	Name:        "path, p",
-	Usage:     "Run app backup",
+	Usage:       "Include path",
-	ArgsUsage: "<domain> [<service>]",
+	Destination: &includePath,
+}
+
+var resticRepo string
+var resticRepoFlag = &cli.StringFlag{
+	Name:        "repo, r",
+	Usage:       "Restic repository",
+	Destination: &resticRepo,
+}
+
+var appBackupListCommand = cli.Command{
+	Name:    "list",
+	Aliases: []string{"ls"},
 	Flags: []cli.Flag{
 		internal.DebugFlag,
 		internal.OfflineFlag,
-		internal.ChaosFlag,
+		snapshotFlag,
+		includePathFlag,
 	},
 	Before:       internal.SubCommandBefore,
+	Usage:        "List all backups",
 	BashComplete: autocomplete.AppNameComplete,
-	Description: `
-Run an app backup.
-
-A backup command and pre/post hook commands are defined in the recipe
-configuration. Abra reads this configuration and run the comands in the context
-of the deployed services. Pass <service> if you only want to back up a single
-service. All backups are placed in the ~/.abra/backups directory.
-
-A single backup file is produced for all backup paths specified for a service.
-If we have the following backup configuration:
-
-    - "backupbot.backup.path=/var/lib/foo,/var/lib/bar"
-
-And we run "abra app backup example.com app", Abra will produce a file that
-looks like:
-
-    ~/.abra/backups/example_com_app_609341138.tar.gz
-
-This file is a compressed archive which contains all backup paths. To see paths, run:
-
-    tar -tf ~/.abra/backups/example_com_app_609341138.tar.gz
-
-(Make sure to change the name of the backup file)
-
-This single file can be used to restore your app. See "abra app restore" for more.
-`,
 	Action: func(c *cli.Context) error {
 		app := internal.ValidateApp(c)
 
-		recipe, err := recipePkg.Get(app.Recipe, internal.Offline)
+		if err := recipe.EnsureExists(app.Recipe); err != nil {
-		if err != nil {
 			logrus.Fatal(err)
 		}
 
 		if !internal.Chaos {
-			if err := recipePkg.EnsureIsClean(app.Recipe); err != nil {
+			if err := recipe.EnsureIsClean(app.Recipe); err != nil {
 				logrus.Fatal(err)
 			}
 
 			if !internal.Offline {
-				if err := recipePkg.EnsureUpToDate(app.Recipe); err != nil {
+				if err := recipe.EnsureUpToDate(app.Recipe); err != nil {
 					logrus.Fatal(err)
 				}
 			}
 
-			if err := recipePkg.EnsureLatest(app.Recipe); err != nil {
+			if err := recipe.EnsureLatest(app.Recipe); err != nil {
 				logrus.Fatal(err)
 			}
 		}
 
-		backupConfigs := make(map[string]backupConfig)
-		for _, service := range recipe.Config.Services {
-			if backupsEnabled, ok := service.Deploy.Labels["backupbot.backup"]; ok {
-				if backupsEnabled == "true" {
-					fullServiceName := fmt.Sprintf("%s_%s", app.StackName(), service.Name)
-					bkConfig := backupConfig{}
-
-					logrus.Debugf("backup config detected for %s", fullServiceName)
-
-					if paths, ok := service.Deploy.Labels["backupbot.backup.path"]; ok {
-						logrus.Debugf("detected backup paths for %s: %s", fullServiceName, paths)
-						bkConfig.backupPaths = strings.Split(paths, ",")
-					}
-
-					if preHookCmd, ok := service.Deploy.Labels["backupbot.backup.pre-hook"]; ok {
-						logrus.Debugf("detected pre-hook command for %s: %s", fullServiceName, preHookCmd)
-						bkConfig.preHookCmd = preHookCmd
-					}
-
-					if postHookCmd, ok := service.Deploy.Labels["backupbot.backup.post-hook"]; ok {
-						logrus.Debugf("detected post-hook command for %s: %s", fullServiceName, postHookCmd)
-						bkConfig.postHookCmd = postHookCmd
-					}
-
-					backupConfigs[service.Name] = bkConfig
-				}
-			}
-		}
-
 		cl, err := client.New(app.Server)
 		if err != nil {
 			logrus.Fatal(err)
 		}
 
-		serviceName := c.Args().Get(1)
+		chosenService, err := service.GetServiceByLabel(context.Background(), cl, config.BackupbotLabel, internal.NoInput)
-		if serviceName != "" {
+		if err != nil {
-			backupConfig, ok := backupConfigs[serviceName]
+			logrus.Fatal(err)
-			if !ok {
+		}
-				logrus.Fatalf("no backup config for %s? does %s exist?", serviceName, serviceName)
-			}
 
-			logrus.Infof("running backup for the %s service", serviceName)
+		logrus.Debugf("retrieved %s as backup enabled service", chosenService.Spec.Name)
 
-			if err := runBackup(cl, app, serviceName, backupConfig); err != nil {
+		filters := filters.NewArgs()
-				logrus.Fatal(err)
+		filters.Add("name", chosenService.Spec.Name)
-			}
+		targetContainer, err := containerPkg.GetContainer(
-		} else {
+			context.Background(),
-			if len(backupConfigs) == 0 {
+			cl,
-				logrus.Fatalf("no backup configs discovered for %s?", app.Name)
+			filters,
-			}
+			internal.NoInput,
+		)
+		if err != nil {
+			logrus.Fatal(err)
+		}
 
-			for serviceName, backupConfig := range backupConfigs {
+		execEnv := []string{fmt.Sprintf("SERVICE=%s", app.Domain)}
-				logrus.Infof("running backup for the %s service", serviceName)
+		if snapshot != "" {
+			logrus.Debugf("including SNAPSHOT=%s in backupbot exec invocation", snapshot)
+			execEnv = append(execEnv, fmt.Sprintf("SNAPSHOT=%s", snapshot))
+		}
+		if includePath != "" {
+			logrus.Debugf("including INCLUDE_PATH=%s in backupbot exec invocation", includePath)
+			execEnv = append(execEnv, fmt.Sprintf("INCLUDE_PATH=%s", includePath))
+		}
 
-				if err := runBackup(cl, app, serviceName, backupConfig); err != nil {
+		execBackupListOpts := types.ExecConfig{
-					logrus.Fatal(err)
+			AttachStderr: true,
-				}
+			AttachStdin:  true,
-			}
+			AttachStdout: true,
+			Cmd:          []string{"/usr/bin/backup", "--", "ls"},
+			Detach:       false,
+			Env:          execEnv,
+			Tty:          true,
+		}
+
+		logrus.Debugf("running backup list on %s with exec config %v", targetContainer.ID, execBackupListOpts)
+
+		// FIXME: avoid instantiating a new CLI
+		dcli, err := command.NewDockerCli()
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execBackupListOpts); err != nil {
+			logrus.Fatal(err)
 		}
 
 		return nil
 	},
 }
 
-// TimeStamp generates a file name friendly timestamp.
+var appBackupDownloadCommand = cli.Command{
-func TimeStamp() string {
+	Name:    "download",
-	ts := time.Now().UTC().Format(time.RFC3339)
+	Aliases: []string{"d"},
-	return strings.Replace(ts, ":", "-", -1)
+	Flags: []cli.Flag{
-}
+		internal.DebugFlag,
+		internal.OfflineFlag,
+		snapshotFlag,
+		includePathFlag,
+	},
+	Before:       internal.SubCommandBefore,
+	Usage:        "Download a backup",
+	BashComplete: autocomplete.AppNameComplete,
+	Action: func(c *cli.Context) error {
+		app := internal.ValidateApp(c)
 
-// runBackup does the actual backup logic.
+		if err := recipe.EnsureExists(app.Recipe); err != nil {
-func runBackup(cl *dockerClient.Client, app config.App, serviceName string, bkConfig backupConfig) error {
+			logrus.Fatal(err)
-	if len(bkConfig.backupPaths) == 0 {
-		return fmt.Errorf("backup paths are empty for %s?", serviceName)
-	}
-
-	// FIXME: avoid instantiating a new CLI
-	dcli, err := command.NewDockerCli()
-	if err != nil {
-		return err
-	}
-
-	filters := filters.NewArgs()
-	filters.Add("name", fmt.Sprintf("^%s_%s", app.StackName(), serviceName))
-
-	targetContainer, err := containerPkg.GetContainer(context.Background(), cl, filters, true)
-	if err != nil {
-		return err
-	}
-
-	fullServiceName := fmt.Sprintf("%s_%s", app.StackName(), serviceName)
-	if bkConfig.preHookCmd != "" {
-		splitCmd := internal.SafeSplit(bkConfig.preHookCmd)
-
-		logrus.Debugf("split pre-hook command for %s into %s", fullServiceName, splitCmd)
-
-		preHookExecOpts := types.ExecConfig{
-			AttachStderr: true,
-			AttachStdin:  true,
-			AttachStdout: true,
-			Cmd:          splitCmd,
-			Detach:       false,
-			Tty:          true,
 		}
 
-		if err := container.RunExec(dcli, cl, targetContainer.ID, &preHookExecOpts); err != nil {
+		if !internal.Chaos {
-			return fmt.Errorf("failed to run %s on %s: %s", bkConfig.preHookCmd, targetContainer.ID, err.Error())
+			if err := recipe.EnsureIsClean(app.Recipe); err != nil {
+				logrus.Fatal(err)
+			}
+
+			if !internal.Offline {
+				if err := recipe.EnsureUpToDate(app.Recipe); err != nil {
+					logrus.Fatal(err)
+				}
+			}
+
+			if err := recipe.EnsureLatest(app.Recipe); err != nil {
+				logrus.Fatal(err)
+			}
 		}
 
-		logrus.Infof("succesfully ran %s pre-hook command: %s", fullServiceName, bkConfig.preHookCmd)
+		cl, err := client.New(app.Server)
-	}
-
-	var tempBackupPaths []string
-	for _, remoteBackupPath := range bkConfig.backupPaths {
-		sanitisedPath := strings.ReplaceAll(remoteBackupPath, "/", "_")
-		localBackupPath := filepath.Join(config.BACKUP_DIR, fmt.Sprintf("%s%s_%s.tar.gz", fullServiceName, sanitisedPath, TimeStamp()))
-		logrus.Debugf("temporarily backing up %s:%s to %s", fullServiceName, remoteBackupPath, localBackupPath)
-
-		logrus.Infof("backing up %s:%s", fullServiceName, remoteBackupPath)
-
-		content, _, err := cl.CopyFromContainer(context.Background(), targetContainer.ID, remoteBackupPath)
 		if err != nil {
-			logrus.Debugf("failed to copy %s from container: %s", remoteBackupPath, err.Error())
+			logrus.Fatal(err)
-			if err := cleanupTempArchives(tempBackupPaths); err != nil {
-				return fmt.Errorf("failed to clean up temporary archives: %s", err.Error())
-			}
-			return fmt.Errorf("failed to copy %s from container: %s", remoteBackupPath, err.Error())
-		}
-		defer content.Close()
-
-		_, srcBase := archive.SplitPathDirEntry(remoteBackupPath)
-		preArchive := archive.RebaseArchiveEntries(content, srcBase, remoteBackupPath)
-		if err := copyToFile(localBackupPath, preArchive); err != nil {
-			logrus.Debugf("failed to create tar archive (%s): %s", localBackupPath, err.Error())
-			if err := cleanupTempArchives(tempBackupPaths); err != nil {
-				return fmt.Errorf("failed to clean up temporary archives: %s", err.Error())
-			}
-			return fmt.Errorf("failed to create tar archive (%s): %s", localBackupPath, err.Error())
 		}
 
-		tempBackupPaths = append(tempBackupPaths, localBackupPath)
+		chosenService, err := service.GetServiceByLabel(context.Background(), cl, config.BackupbotLabel, internal.NoInput)
-	}
+		if err != nil {
-
+			logrus.Fatal(err)
-	logrus.Infof("compressing and merging archives...")
-
-	if err := mergeArchives(tempBackupPaths, fullServiceName); err != nil {
-		logrus.Debugf("failed to merge archive files: %s", err.Error())
-		if err := cleanupTempArchives(tempBackupPaths); err != nil {
-			return fmt.Errorf("failed to clean up temporary archives: %s", err.Error())
 		}
-		return fmt.Errorf("failed to merge archive files: %s", err.Error())
-	}
 
-	if err := cleanupTempArchives(tempBackupPaths); err != nil {
+		logrus.Debugf("retrieved %s as backup enabled service", chosenService.Spec.Name)
-		return fmt.Errorf("failed to clean up temporary archives: %s", err.Error())
-	}
 
-	if bkConfig.postHookCmd != "" {
+		filters := filters.NewArgs()
-		splitCmd := internal.SafeSplit(bkConfig.postHookCmd)
+		filters.Add("name", chosenService.Spec.Name)
+		targetContainer, err := containerPkg.GetContainer(
+			context.Background(),
+			cl,
+			filters,
+			internal.NoInput,
+		)
+		if err != nil {
+			logrus.Fatal(err)
+		}
 
-		logrus.Debugf("split post-hook command for %s into %s", fullServiceName, splitCmd)
+		execEnv := []string{fmt.Sprintf("SERVICE=%s", app.Domain)}
+		if snapshot != "" {
+			logrus.Debugf("including SNAPSHOT=%s in backupbot exec invocation", snapshot)
+			execEnv = append(execEnv, fmt.Sprintf("SNAPSHOT=%s", snapshot))
+		}
+		if includePath != "" {
+			logrus.Debugf("including INCLUDE_PATH=%s in backupbot exec invocation", includePath)
+			execEnv = append(execEnv, fmt.Sprintf("INCLUDE_PATH=%s", includePath))
+		}
 
-		postHookExecOpts := types.ExecConfig{
+		execBackupListOpts := types.ExecConfig{
 			AttachStderr: true,
 			AttachStdin:  true,
 			AttachStdout: true,
-			Cmd:          splitCmd,
+			Cmd:          []string{"/usr/bin/backup", "--", "download"},
 			Detach:       false,
+			Env:          execEnv,
 			Tty:          true,
 		}
 
-		if err := container.RunExec(dcli, cl, targetContainer.ID, &postHookExecOpts); err != nil {
+		logrus.Debugf("running backup list on %s with exec config %v", targetContainer.ID, execBackupListOpts)
-			return err
+
+		// FIXME: avoid instantiating a new CLI
+		dcli, err := command.NewDockerCli()
+		if err != nil {
+			logrus.Fatal(err)
 		}
 
-		logrus.Infof("succesfully ran %s post-hook command: %s", fullServiceName, bkConfig.postHookCmd)
+		out, err := container.RunExec(dcli, cl, targetContainer.ID, &execBackupListOpts)
-	}
+		if err != nil {
-
+			logrus.Fatal(err)
-	return nil
-}
-
-func copyToFile(outfile string, r io.Reader) error {
-	tmpFile, err := os.CreateTemp(filepath.Dir(outfile), ".tar_temp")
-	if err != nil {
-		return err
-	}
-
-	tmpPath := tmpFile.Name()
-
-	_, err = io.Copy(tmpFile, r)
-	tmpFile.Close()
-
-	if err != nil {
-		os.Remove(tmpPath)
-		return err
-	}
-
-	if err = os.Rename(tmpPath, outfile); err != nil {
-		os.Remove(tmpPath)
-		return err
-	}
-
-	return nil
-}
-
-func cleanupTempArchives(tarPaths []string) error {
-	for _, tarPath := range tarPaths {
-		if err := os.RemoveAll(tarPath); err != nil {
-			return err
 		}
 
-		logrus.Debugf("remove temporary archive file %s", tarPath)
+		fmt.Println("OUTPUT: ", out)
-	}
 
-	return nil
+		return nil
+	},
 }
 
-func mergeArchives(tarPaths []string, serviceName string) error {
+var appBackupCreateCommand = cli.Command{
-	var out io.Writer
+	Name:    "create",
-	var cout *pgzip.Writer
+	Aliases: []string{"c"},
+	Flags: []cli.Flag{
+		internal.DebugFlag,
+		internal.OfflineFlag,
+		resticRepoFlag,
+	},
+	Before:       internal.SubCommandBefore,
+	Usage:        "Create a new backup",
+	BashComplete: autocomplete.AppNameComplete,
+	Action: func(c *cli.Context) error {
+		app := internal.ValidateApp(c)
 
-	localBackupPath := filepath.Join(config.BACKUP_DIR, fmt.Sprintf("%s_%s.tar.gz", serviceName, TimeStamp()))
+		if err := recipe.EnsureExists(app.Recipe); err != nil {
-
+			logrus.Fatal(err)
-	fout, err := os.Create(localBackupPath)
-	if err != nil {
-		return fmt.Errorf("Failed to open %s: %s", localBackupPath, err)
-	}
-
-	defer fout.Close()
-	out = fout
-
-	cout = pgzip.NewWriter(out)
-	out = cout
-
-	tw := tar.NewWriter(out)
-
-	for _, tarPath := range tarPaths {
-		if err := addTar(tw, tarPath); err != nil {
-			return fmt.Errorf("failed to merge %s: %v", tarPath, err)
 		}
-	}
 
-	if err := tw.Close(); err != nil {
+		if !internal.Chaos {
-		return fmt.Errorf("failed to close tar writer %v", err)
+			if err := recipe.EnsureIsClean(app.Recipe); err != nil {
-	}
+				logrus.Fatal(err)
-
+			}
-	if cout != nil {
+
-		if err := cout.Flush(); err != nil {
+			if !internal.Offline {
-			return fmt.Errorf("failed to flush: %s", err)
+				if err := recipe.EnsureUpToDate(app.Recipe); err != nil {
-		} else if err = cout.Close(); err != nil {
+					logrus.Fatal(err)
-			return fmt.Errorf("failed to close compressed writer: %s", err)
+				}
-		}
+			}
-	}
+
-
+			if err := recipe.EnsureLatest(app.Recipe); err != nil {
-	logrus.Infof("backed up %s to %s", serviceName, localBackupPath)
+				logrus.Fatal(err)
-
-	return nil
-}
-
-func addTar(tw *tar.Writer, pth string) (err error) {
-	var tr *tar.Reader
-	var rc io.ReadCloser
-	var hdr *tar.Header
-
-	if tr, rc, err = openTarFile(pth); err != nil {
-		return
-	}
-
-	for {
-		if hdr, err = tr.Next(); err != nil {
-			if err == io.EOF {
-				err = nil
 			}
-			break
 		}
-		if err = tw.WriteHeader(hdr); err != nil {
+
-			break
+		cl, err := client.New(app.Server)
-		} else if _, err = io.Copy(tw, tr); err != nil {
+		if err != nil {
-			break
+			logrus.Fatal(err)
 		}
-	}
+
-	if err == nil {
+		chosenService, err := service.GetServiceByLabel(context.Background(), cl, config.BackupbotLabel, internal.NoInput)
-		err = rc.Close()
+		if err != nil {
-	} else {
+			logrus.Fatal(err)
-		rc.Close()
+		}
-	}
+
-	return
+		logrus.Debugf("retrieved %s as backup enabled service", chosenService.Spec.Name)
+
+		filters := filters.NewArgs()
+		filters.Add("name", chosenService.Spec.Name)
+		targetContainer, err := containerPkg.GetContainer(
+			context.Background(),
+			cl,
+			filters,
+			internal.NoInput,
+		)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		execEnv := []string{fmt.Sprintf("SERVICE=%s", app.Domain)}
+		if resticRepo != "" {
+			logrus.Debugf("including RESTIC_REPO=%s in backupbot exec invocation", resticRepo)
+			execEnv = append(execEnv, fmt.Sprintf("RESTIC_REPO=%s", resticRepo))
+		}
+		execBackupListOpts := types.ExecConfig{
+			AttachStderr: true,
+			AttachStdin:  true,
+			AttachStdout: true,
+			Cmd:          []string{"/usr/bin/backup", "--", "create"},
+			Detach:       false,
+			Env:          execEnv,
+			Tty:          true,
+		}
+
+		logrus.Debugf("running backup list on %s with exec config %v", targetContainer.ID, execBackupListOpts)
+
+		// FIXME: avoid instantiating a new CLI
+		dcli, err := command.NewDockerCli()
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execBackupListOpts); err != nil {
+			logrus.Fatal(err)
+		}
+
+		return nil
+	},
 }
 
-func openTarFile(pth string) (tr *tar.Reader, rc io.ReadCloser, err error) {
+var appBackupSnapshotsCommand = cli.Command{
-	var fin *os.File
+	Name:    "snapshots",
-	var n int
+	Aliases: []string{"s"},
-	buff := make([]byte, 1024)
+	Flags: []cli.Flag{
+		internal.DebugFlag,
+		internal.OfflineFlag,
+		snapshotFlag,
+	},
+	Before:       internal.SubCommandBefore,
+	Usage:        "List backup snapshots",
+	BashComplete: autocomplete.AppNameComplete,
+	Action: func(c *cli.Context) error {
+		app := internal.ValidateApp(c)
 
-	if fin, err = os.Open(pth); err != nil {
+		if err := recipe.EnsureExists(app.Recipe); err != nil {
-		return
+			logrus.Fatal(err)
-	}
+		}
 
-	if n, err = fin.Read(buff); err != nil {
+		if !internal.Chaos {
-		fin.Close()
+			if err := recipe.EnsureIsClean(app.Recipe); err != nil {
-		return
+				logrus.Fatal(err)
-	} else if n == 0 {
+			}
-		fin.Close()
-		err = fmt.Errorf("%s is empty", pth)
-		return
-	}
 
-	if _, err = fin.Seek(0, 0); err != nil {
+			if !internal.Offline {
-		fin.Close()
+				if err := recipe.EnsureUpToDate(app.Recipe); err != nil {
-		return
+					logrus.Fatal(err)
-	}
+				}
+			}
 
-	rc = fin
+			if err := recipe.EnsureLatest(app.Recipe); err != nil {
-	tr = tar.NewReader(rc)
+				logrus.Fatal(err)
+			}
+		}
 
-	return tr, rc, nil
+		cl, err := client.New(app.Server)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		chosenService, err := service.GetServiceByLabel(context.Background(), cl, config.BackupbotLabel, internal.NoInput)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		logrus.Debugf("retrieved %s as backup enabled service", chosenService.Spec.Name)
+
+		filters := filters.NewArgs()
+		filters.Add("name", chosenService.Spec.Name)
+		targetContainer, err := containerPkg.GetContainer(
+			context.Background(),
+			cl,
+			filters,
+			internal.NoInput,
+		)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		execEnv := []string{fmt.Sprintf("SERVICE=%s", app.Domain)}
+		if snapshot != "" {
+			logrus.Debugf("including SNAPSHOT=%s in backupbot exec invocation", snapshot)
+			execEnv = append(execEnv, fmt.Sprintf("SNAPSHOT=%s", snapshot))
+		}
+		execBackupListOpts := types.ExecConfig{
+			AttachStderr: true,
+			AttachStdin:  true,
+			AttachStdout: true,
+			Cmd:          []string{"/usr/bin/backup", "--", "snapshots"},
+			Detach:       false,
+			Env:          execEnv,
+			Tty:          true,
+		}
+
+		logrus.Debugf("running backup list on %s with exec config %v", targetContainer.ID, execBackupListOpts)
+
+		// FIXME: avoid instantiating a new CLI
+		dcli, err := command.NewDockerCli()
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execBackupListOpts); err != nil {
+			logrus.Fatal(err)
+		}
+
+		return nil
+	},
+}
+
+var appBackupCommand = cli.Command{
+	Name:      "backup",
+	Aliases:   []string{"b"},
+	Usage:     "Manage app backups",
+	ArgsUsage: "<domain>",
+	Subcommands: []cli.Command{
+		appBackupListCommand,
+		appBackupSnapshotsCommand,
+		appBackupDownloadCommand,
+		appBackupCreateCommand,
+	},
 }

cli/app/restore.go

@@ -2,9 +2,7 @@ package app
 
 import (
 	"context"
-	"errors"
 	"fmt"
-	"os"
 
 	"coopcloud.tech/abra/cli/internal"
 	"coopcloud.tech/abra/pkg/autocomplete"
@@ -12,212 +10,112 @@ import (
 	"coopcloud.tech/abra/pkg/config"
 	containerPkg "coopcloud.tech/abra/pkg/container"
 	"coopcloud.tech/abra/pkg/recipe"
-	recipePkg "coopcloud.tech/abra/pkg/recipe"
+	"coopcloud.tech/abra/pkg/service"
 	"coopcloud.tech/abra/pkg/upstream/container"
 	"github.com/docker/cli/cli/command"
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/filters"
-	dockerClient "github.com/docker/docker/client"
-	"github.com/docker/docker/pkg/archive"
 	"github.com/sirupsen/logrus"
 	"github.com/urfave/cli"
 )
 
-type restoreConfig struct {
+var targetPath string
-	preHookCmd  string
+var targetPathFlag = &cli.StringFlag{
-	postHookCmd string
+	Name:        "target, t",
+	Usage:       "Target path",
+	Destination: &targetPath,
 }
 
 var appRestoreCommand = cli.Command{
 	Name:      "restore",
 	Aliases:   []string{"rs"},
-	Usage:     "Run app restore",
+	Usage:     "Restore an app backup",
-	ArgsUsage: "<domain> <service> <file>",
+	ArgsUsage: "<domain> <service>",
 	Flags: []cli.Flag{
 		internal.DebugFlag,
 		internal.OfflineFlag,
-		internal.ChaosFlag,
+		targetPathFlag,
 	},
 	Before:       internal.SubCommandBefore,
 	BashComplete: autocomplete.AppNameComplete,
-	Description: `
-Run an app restore.
-
-Pre/post hook commands are defined in the recipe configuration. Abra reads this
-configuration and run the comands in the context of the service before
-restoring the backup.
-
-Unlike "abra app backup", restore must be run on a per-service basis. You can
-not restore all services in one go. Backup files produced by Abra are
-compressed archives which use absolute paths. This allows Abra to restore
-according to standard tar command logic, i.e. the backup will be restored to
-the path it was originally backed up from.
-
-Example:
-
-    abra app restore example.com app ~/.abra/backups/example_com_app_609341138.tar.gz
-`,
 	Action: func(c *cli.Context) error {
 		app := internal.ValidateApp(c)
 
-		recipe, err := recipe.Get(app.Recipe, internal.Offline)
+		if err := recipe.EnsureExists(app.Recipe); err != nil {
-		if err != nil {
 			logrus.Fatal(err)
 		}
 
 		if !internal.Chaos {
-			if err := recipePkg.EnsureIsClean(app.Recipe); err != nil {
+			if err := recipe.EnsureIsClean(app.Recipe); err != nil {
 				logrus.Fatal(err)
 			}
 
 			if !internal.Offline {
-				if err := recipePkg.EnsureUpToDate(app.Recipe); err != nil {
+				if err := recipe.EnsureUpToDate(app.Recipe); err != nil {
 					logrus.Fatal(err)
 				}
 			}
 
-			if err := recipePkg.EnsureLatest(app.Recipe); err != nil {
+			if err := recipe.EnsureLatest(app.Recipe); err != nil {
 				logrus.Fatal(err)
 			}
 		}
 
-		serviceName := c.Args().Get(1)
-		if serviceName == "" {
-			internal.ShowSubcommandHelpAndError(c, errors.New("missing <service>?"))
-		}
-
-		backupPath := c.Args().Get(2)
-		if backupPath == "" {
-			internal.ShowSubcommandHelpAndError(c, errors.New("missing <file>?"))
-		}
-
-		if _, err := os.Stat(backupPath); err != nil {
-			if os.IsNotExist(err) {
-				logrus.Fatalf("%s doesn't exist?", backupPath)
-			}
-		}
-
-		restoreConfigs := make(map[string]restoreConfig)
-		for _, service := range recipe.Config.Services {
-			if restoreEnabled, ok := service.Deploy.Labels["backupbot.restore"]; ok {
-				if restoreEnabled == "true" {
-					fullServiceName := fmt.Sprintf("%s_%s", app.StackName(), service.Name)
-					rsConfig := restoreConfig{}
-
-					logrus.Debugf("restore config detected for %s", fullServiceName)
-
-					if preHookCmd, ok := service.Deploy.Labels["backupbot.restore.pre-hook"]; ok {
-						logrus.Debugf("detected pre-hook command for %s: %s", fullServiceName, preHookCmd)
-						rsConfig.preHookCmd = preHookCmd
-					}
-
-					if postHookCmd, ok := service.Deploy.Labels["backupbot.restore.post-hook"]; ok {
-						logrus.Debugf("detected post-hook command for %s: %s", fullServiceName, postHookCmd)
-						rsConfig.postHookCmd = postHookCmd
-					}
-
-					restoreConfigs[service.Name] = rsConfig
-				}
-			}
-		}
-
-		rsConfig, ok := restoreConfigs[serviceName]
-		if !ok {
-			rsConfig = restoreConfig{}
-		}
-
 		cl, err := client.New(app.Server)
 		if err != nil {
 			logrus.Fatal(err)
 		}
 
-		if err := runRestore(cl, app, backupPath, serviceName, rsConfig); err != nil {
+		chosenService, err := service.GetServiceByLabel(context.Background(), cl, config.BackupbotLabel, internal.NoInput)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		logrus.Debugf("retrieved %s as backup enabled service", chosenService.Spec.Name)
+
+		filters := filters.NewArgs()
+		filters.Add("name", chosenService.Spec.Name)
+		targetContainer, err := containerPkg.GetContainer(
+			context.Background(),
+			cl,
+			filters,
+			internal.NoInput,
+		)
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		execEnv := []string{fmt.Sprintf("SERVICE=%s", app.Domain)}
+		if snapshot != "" {
+			logrus.Debugf("including SNAPSHOT=%s in backupbot exec invocation", snapshot)
+			execEnv = append(execEnv, fmt.Sprintf("SNAPSHOT=%s", snapshot))
+		}
+		if targetPath != "" {
+			logrus.Debugf("including TARGET=%s in backupbot exec invocation", targetPath)
+			execEnv = append(execEnv, fmt.Sprintf("TARGET=%s", targetPath))
+		}
+		execBackupListOpts := types.ExecConfig{
+			AttachStderr: true,
+			AttachStdin:  true,
+			AttachStdout: true,
+			Cmd:          []string{"/usr/bin/backup", "--", "restore"},
+			Detach:       false,
+			Env:          execEnv,
+			Tty:          true,
+		}
+
+		logrus.Debugf("running backup list on %s with exec config %v", targetContainer.ID, execBackupListOpts)
+
+		// FIXME: avoid instantiating a new CLI
+		dcli, err := command.NewDockerCli()
+		if err != nil {
+			logrus.Fatal(err)
+		}
+
+		if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execBackupListOpts); err != nil {
 			logrus.Fatal(err)
 		}
 
 		return nil
 	},
 }
-
-// runRestore does the actual restore logic.
-func runRestore(cl *dockerClient.Client, app config.App, backupPath, serviceName string, rsConfig restoreConfig) error {
-	// FIXME: avoid instantiating a new CLI
-	dcli, err := command.NewDockerCli()
-	if err != nil {
-		return err
-	}
-
-	filters := filters.NewArgs()
-	filters.Add("name", fmt.Sprintf("^%s_%s", app.StackName(), serviceName))
-
-	targetContainer, err := containerPkg.GetContainer(context.Background(), cl, filters, true)
-	if err != nil {
-		return err
-	}
-
-	fullServiceName := fmt.Sprintf("%s_%s", app.StackName(), serviceName)
-	if rsConfig.preHookCmd != "" {
-		splitCmd := internal.SafeSplit(rsConfig.preHookCmd)
-
-		logrus.Debugf("split pre-hook command for %s into %s", fullServiceName, splitCmd)
-
-		preHookExecOpts := types.ExecConfig{
-			AttachStderr: true,
-			AttachStdin:  true,
-			AttachStdout: true,
-			Cmd:          splitCmd,
-			Detach:       false,
-			Tty:          true,
-		}
-
-		if err := container.RunExec(dcli, cl, targetContainer.ID, &preHookExecOpts); err != nil {
-			return err
-		}
-
-		logrus.Infof("succesfully ran %s pre-hook command: %s", fullServiceName, rsConfig.preHookCmd)
-	}
-
-	backupReader, err := os.Open(backupPath)
-	if err != nil {
-		return err
-	}
-
-	content, err := archive.DecompressStream(backupReader)
-	if err != nil {
-		return err
-	}
-
-	// NOTE(d1): we use absolute paths so tar knows what to do. it will restore
-	// files according to the paths set in the compressed archive
-	restorePath := "/"
-
-	copyOpts := types.CopyToContainerOptions{AllowOverwriteDirWithFile: false, CopyUIDGID: false}
-	if err := cl.CopyToContainer(context.Background(), targetContainer.ID, restorePath, content, copyOpts); err != nil {
-		return err
-	}
-
-	logrus.Infof("restored %s to %s", backupPath, fullServiceName)
-
-	if rsConfig.postHookCmd != "" {
-		splitCmd := internal.SafeSplit(rsConfig.postHookCmd)
-
-		logrus.Debugf("split post-hook command for %s into %s", fullServiceName, splitCmd)
-
-		postHookExecOpts := types.ExecConfig{
-			AttachStderr: true,
-			AttachStdin:  true,
-			AttachStdout: true,
-			Cmd:          splitCmd,
-			Detach:       false,
-			Tty:          true,
-		}
-
-		if err := container.RunExec(dcli, cl, targetContainer.ID, &postHookExecOpts); err != nil {
-			return err
-		}
-
-		logrus.Infof("succesfully ran %s post-hook command: %s", fullServiceName, rsConfig.postHookCmd)
-	}
-
-	return nil
-}

cli/app/run.go

@@ -91,7 +91,7 @@ var appRunCommand = cli.Command{
 			logrus.Fatal(err)
 		}
 
-		if err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
+		if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
 			logrus.Fatal(err)
 		}
 

cli/internal/command.go

@@ -60,7 +60,7 @@ func RunCmdRemote(cl *dockerClient.Client, app config.App, abraSh, serviceName,
 		Tty:          false,
 	}
 
-	if err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
+	if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
 		logrus.Infof("%s does not exist for %s, use /bin/sh as fallback", shell, app.Name)
 		shell = "/bin/sh"
 	}
@@ -85,7 +85,7 @@ func RunCmdRemote(cl *dockerClient.Client, app config.App, abraSh, serviceName,
 		execCreateOpts.Tty = false
 	}
 
-	if err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
+	if _, err := container.RunExec(dcli, cl, targetContainer.ID, &execCreateOpts); err != nil {
 		return err
 	}
 

pkg/config/env.go

@@ -36,6 +36,8 @@ var REPOS_BASE_URL = "https://git.coopcloud.tech/coop-cloud"
 var CATALOGUE_JSON_REPO_NAME = "recipes-catalogue-json"
 var SSH_URL_TEMPLATE = "ssh://git@git.coopcloud.tech:2222/coop-cloud/%s.git"
 
+var BackupbotLabel = "coop-cloud.backupbot.enabled"
+
 // envVarModifiers is a list of env var modifier strings. These are added to
 // env vars as comments and modify their processing by Abra, e.g. determining
 // how long secrets should be.

pkg/container/container.go

@@ -28,7 +28,7 @@ func GetContainer(c context.Context, cl *client.Client, filters filters.Args, no
 		return types.Container{}, fmt.Errorf("no containers matching the %v filter found?", filter)
 	}
 
-	if len(containers) != 1 {
+	if len(containers) > 1 {
 		var containersRaw []string
 		for _, container := range containers {
 			containerName := strings.Join(container.Names, " ")

pkg/service/service.go

@@ -14,6 +14,69 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
+// GetService retrieves a service container based on a label. If prompt is true
+// and the retrievd count of service containers does not match 1, then a prompt
+// is presented to let the user choose. A count of 0 is handled gracefully.
+func GetServiceByLabel(c context.Context, cl *client.Client, label string, prompt bool) (swarm.Service, error) {
+	services, err := cl.ServiceList(c, types.ServiceListOptions{})
+	if err != nil {
+		return swarm.Service{}, err
+	}
+
+	if len(services) == 0 {
+		return swarm.Service{}, fmt.Errorf("no services deployed?")
+	}
+
+	var backupServices []swarm.Service
+	for _, service := range services {
+		if enabled, exists := service.Spec.Labels[label]; exists && enabled == "true" {
+			backupServices = append(backupServices, service)
+		}
+	}
+
+	if len(backupServices) == 0 {
+		return swarm.Service{}, fmt.Errorf("no backup services deployed?")
+	}
+
+	if len(backupServices) > 1 {
+		var servicesRaw []string
+		for _, service := range backupServices {
+			serviceName := service.Spec.Name
+			created := formatter.HumanDuration(service.CreatedAt.Unix())
+			servicesRaw = append(servicesRaw, fmt.Sprintf("%s (created %v)", serviceName, created))
+		}
+
+		if !prompt {
+			err := fmt.Errorf("expected 1 service but found %v: %s", len(backupServices), strings.Join(servicesRaw, " "))
+			return swarm.Service{}, err
+		}
+
+		logrus.Warnf("ambiguous service list received, prompting for input")
+
+		var response string
+		prompt := &survey.Select{
+			Message: "which service are you looking for?",
+			Options: servicesRaw,
+		}
+
+		if err := survey.AskOne(prompt, &response); err != nil {
+			return swarm.Service{}, err
+		}
+
+		chosenService := strings.TrimSpace(strings.Split(response, " ")[0])
+		for _, service := range backupServices {
+			serviceName := strings.ToLower(service.Spec.Name)
+			if serviceName == chosenService {
+				return service, nil
+			}
+		}
+
+		logrus.Panic("failed to match chosen service")
+	}
+
+	return backupServices[0], nil
+}
+
 // GetService retrieves a service container. If prompt is true and the retrievd
 // count of service containers does not match 1, then a prompt is presented to
 // let the user choose. A count of 0 is handled gracefully.

pkg/upstream/container/exec.go

@@ -13,7 +13,8 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
-func RunExec(dockerCli command.Cli, client *apiclient.Client, containerID string, execConfig *types.ExecConfig) error {
+func RunExec(dockerCli command.Cli, client *apiclient.Client, containerID string,
+	execConfig *types.ExecConfig) (io.Writer, error) {
 	ctx := context.Background()
 
 	// We need to check the tty _before_ we do the ContainerExecCreate, because
@@ -21,22 +22,22 @@ func RunExec(dockerCli command.Cli, client *apiclient.Client, containerID string
 	// there's no easy way to clean those up). But also in order to make "not
 	// exist" errors take precedence we do a dummy inspect first.
 	if _, err := client.ContainerInspect(ctx, containerID); err != nil {
-		return err
+		return nil, err
 	}
 	if !execConfig.Detach {
 		if err := dockerCli.In().CheckTty(execConfig.AttachStdin, execConfig.Tty); err != nil {
-			return err
+			return nil, err
 		}
 	}
 
 	response, err := client.ContainerExecCreate(ctx, containerID, *execConfig)
 	if err != nil {
-		return err
+		return nil, err
 	}
 
 	execID := response.ID
 	if execID == "" {
-		return errors.New("exec ID empty")
+		return nil, errors.New("exec ID empty")
 	}
 
 	if execConfig.Detach {
@@ -44,13 +45,13 @@ func RunExec(dockerCli command.Cli, client *apiclient.Client, containerID string
 			Detach: execConfig.Detach,
 			Tty:    execConfig.Tty,
 		}
-		return client.ContainerExecStart(ctx, execID, execStartCheck)
+		return nil, client.ContainerExecStart(ctx, execID, execStartCheck)
 	}
 	return interactiveExec(ctx, dockerCli, client, execConfig, execID)
 }
 
 func interactiveExec(ctx context.Context, dockerCli command.Cli, client *apiclient.Client,
-	execConfig *types.ExecConfig, execID string) error {
+	execConfig *types.ExecConfig, execID string) (io.Writer, error) {
 	// Interactive exec requested.
 	var (
 		out, stderr io.Writer
@@ -76,7 +77,7 @@ func interactiveExec(ctx context.Context, dockerCli command.Cli, client *apiclie
 	}
 	resp, err := client.ContainerExecAttach(ctx, execID, execStartCheck)
 	if err != nil {
-		return err
+		return out, err
 	}
 	defer resp.Close()
 
@@ -107,10 +108,10 @@ func interactiveExec(ctx context.Context, dockerCli command.Cli, client *apiclie
 
 	if err := <-errCh; err != nil {
 		logrus.Debugf("Error hijack: %s", err)
-		return err
+		return out, err
 	}
 
-	return getExecExitStatus(ctx, client, execID)
+	return out, getExecExitStatus(ctx, client, execID)
 }
 
 func getExecExitStatus(ctx context.Context, client apiclient.ContainerAPIClient, execID string) error {