docker-volume-backup/cmd/backup/main.go

509 lines
14 KiB
Go
Raw Normal View History

2021-08-22 18:07:32 +02:00
// Copyright 2021 - Offen Authors <hioffen@posteo.de>
// SPDX-License-Identifier: MPL-2.0
package main
import (
2021-08-22 14:44:33 +02:00
"bytes"
"context"
2021-08-21 19:26:42 +02:00
"errors"
"fmt"
2021-08-21 21:26:27 +02:00
"io"
2021-08-22 14:44:33 +02:00
"io/ioutil"
"os"
2021-08-21 21:26:27 +02:00
"path"
2021-08-22 14:00:21 +02:00
"path/filepath"
"strconv"
"time"
"github.com/docker/docker/api/types"
"github.com/docker/docker/api/types/filters"
"github.com/docker/docker/api/types/swarm"
"github.com/docker/docker/client"
"github.com/joho/godotenv"
2021-08-22 21:06:51 +02:00
"github.com/leekchan/timeutil"
2021-08-21 21:26:27 +02:00
minio "github.com/minio/minio-go/v7"
"github.com/minio/minio-go/v7/pkg/credentials"
2021-08-22 16:41:06 +02:00
"github.com/sirupsen/logrus"
2021-08-21 21:26:27 +02:00
"github.com/walle/targz"
2021-08-22 14:44:33 +02:00
"golang.org/x/crypto/openpgp"
)
func main() {
2021-08-22 19:37:48 +02:00
unlock := lock("/var/dockervolumebackup.lock")
defer unlock()
2021-08-21 19:26:42 +02:00
s := &script{}
2021-08-22 19:37:48 +02:00
s.must(s.init())
s.must(s.stopContainersAndRun(s.takeBackup))
s.must(s.encryptBackup())
s.must(s.copyBackup())
s.must(s.cleanBackup())
s.must(s.pruneOldBackups())
s.logger.Info("Finished running backup tasks.")
2021-08-21 19:26:42 +02:00
}
type script struct {
2021-08-22 22:02:19 +02:00
ctx context.Context
cli *client.Client
mc *minio.Client
logger *logrus.Logger
start time.Time
file string
bucket string
archive string
sources string
passphrase []byte
retentionDays *int
leeway *time.Duration
containerLabel string
pruningPrefix string
2021-08-21 19:26:42 +02:00
}
2021-08-22 19:37:48 +02:00
// lock opens a lockfile at the given location, keeping it locked until the
// caller invokes the returned release func. When invoked while the file is
// still locked the function panics.
func lock(lockfile string) func() error {
lf, err := os.OpenFile(lockfile, os.O_CREATE, os.ModeAppend)
if err != nil {
2021-08-22 19:37:48 +02:00
panic(err)
}
return func() error {
if err := lf.Close(); err != nil {
return fmt.Errorf("lock: error releasing file lock: %w", err)
}
if err := os.Remove(lockfile); err != nil {
return fmt.Errorf("lock: error removing lock file: %w", err)
}
return nil
2021-08-22 19:37:48 +02:00
}
}
2021-08-21 19:26:42 +02:00
// init creates all resources needed for the script to perform actions against
// remote resources like the Docker engine or remote storage locations.
2021-08-21 19:26:42 +02:00
func (s *script) init() error {
s.ctx = context.Background()
2021-08-22 16:41:06 +02:00
s.logger = logrus.New()
s.logger.SetOutput(os.Stdout)
2021-08-21 19:26:42 +02:00
if err := godotenv.Load("/etc/backup.env"); err != nil {
return fmt.Errorf("init: failed to load env file: %w", err)
}
_, err := os.Stat("/var/run/docker.sock")
if !os.IsNotExist(err) {
2021-08-21 19:26:42 +02:00
cli, err := client.NewClientWithOpts(client.FromEnv, client.WithAPIVersionNegotiation())
if err != nil {
2021-08-22 16:41:06 +02:00
return fmt.Errorf("init: failed to create docker client")
}
2021-08-21 19:26:42 +02:00
s.cli = cli
}
2021-08-21 21:26:27 +02:00
if bucket := os.Getenv("AWS_S3_BUCKET_NAME"); bucket != "" {
s.bucket = bucket
2021-08-21 21:26:27 +02:00
mc, err := minio.New(os.Getenv("AWS_ENDPOINT"), &minio.Options{
Creds: credentials.NewStaticV4(
os.Getenv("AWS_ACCESS_KEY_ID"),
os.Getenv("AWS_SECRET_ACCESS_KEY"),
"",
),
2021-08-22 19:26:34 +02:00
Secure: os.Getenv("AWS_ENDPOINT_INSECURE") == "" && os.Getenv("AWS_ENDPOINT_PROTO") == "https",
2021-08-21 21:26:27 +02:00
})
if err != nil {
return fmt.Errorf("init: error setting up minio client: %w", err)
}
s.mc = mc
}
2021-08-22 16:41:06 +02:00
file := os.Getenv("BACKUP_FILENAME")
if file == "" {
return errors.New("init: BACKUP_FILENAME not given")
}
s.file = path.Join("/tmp", file)
2021-08-22 16:41:06 +02:00
s.archive = os.Getenv("BACKUP_ARCHIVE")
2021-08-22 19:37:48 +02:00
s.sources = os.Getenv("BACKUP_SOURCES")
2021-08-22 22:02:19 +02:00
if v := os.Getenv("GPG_PASSPHRASE"); v != "" {
s.passphrase = []byte(v)
}
if v := os.Getenv("BACKUP_RETENTION_DAYS"); v != "" {
i, err := strconv.Atoi(v)
if err != nil {
return fmt.Errorf("init: error parsing BACKUP_RETENTION_DAYS as int: %w", err)
}
s.retentionDays = &i
}
if v := os.Getenv("BACKUP_PRUNING_LEEWAY"); v != "" {
d, err := time.ParseDuration(v)
if err != nil {
return fmt.Errorf("init: error parsing BACKUP_PRUNING_LEEWAY as duration: %w", err)
}
s.leeway = &d
}
s.containerLabel = os.Getenv("BACKUP_STOP_CONTAINER_LABEL")
s.pruningPrefix = os.Getenv("BACKUP_PRUNING_PREFIX")
s.start = time.Now()
2021-08-22 21:06:51 +02:00
2021-08-21 19:26:42 +02:00
return nil
}
// stopContainersAndRun stops all Docker containers that are marked as to being
// stopped during the backup and runs the given thunk. After returning, it makes
// sure containers are being restarted if required.
func (s *script) stopContainersAndRun(thunk func() error) error {
2021-08-21 19:26:42 +02:00
if s.cli == nil {
return thunk()
2021-08-21 19:26:42 +02:00
}
allContainers, err := s.cli.ContainerList(s.ctx, types.ContainerListOptions{
Quiet: true,
})
if err != nil {
return fmt.Errorf("stopContainersAndRun: error querying for containers: %w", err)
}
2021-08-22 19:37:48 +02:00
containerLabel := fmt.Sprintf(
"docker-volume-backup.stop-during-backup=%s",
2021-08-22 22:02:19 +02:00
s.containerLabel,
2021-08-22 19:37:48 +02:00
)
containersToStop, err := s.cli.ContainerList(s.ctx, types.ContainerListOptions{
2021-08-21 19:26:42 +02:00
Quiet: true,
Filters: filters.NewArgs(filters.KeyValuePair{
2021-08-22 19:37:48 +02:00
Key: "label",
Value: containerLabel,
2021-08-21 19:26:42 +02:00
}),
})
if err != nil {
return fmt.Errorf("stopContainersAndRun: error querying for containers to stop: %w", err)
}
2021-08-22 19:37:48 +02:00
s.logger.Infof(
2021-08-22 22:02:19 +02:00
"Stopping %d containers labeled `%s` out of %d running container(s).",
2021-08-22 19:37:48 +02:00
len(containersToStop),
containerLabel,
len(allContainers),
)
var stoppedContainers []types.Container
var errors []error
if len(containersToStop) != 0 {
for _, container := range containersToStop {
2021-08-21 19:26:42 +02:00
if err := s.cli.ContainerStop(s.ctx, container.ID, nil); err != nil {
errors = append(errors, err)
} else {
stoppedContainers = append(stoppedContainers, container)
}
}
}
defer func() error {
servicesRequiringUpdate := map[string]struct{}{}
var restartErrors []error
for _, container := range stoppedContainers {
if swarmServiceName, ok := container.Labels["com.docker.swarm.service.name"]; ok {
servicesRequiringUpdate[swarmServiceName] = struct{}{}
continue
}
if err := s.cli.ContainerStart(s.ctx, container.ID, types.ContainerStartOptions{}); err != nil {
restartErrors = append(restartErrors, err)
}
}
if len(servicesRequiringUpdate) != 0 {
services, _ := s.cli.ServiceList(s.ctx, types.ServiceListOptions{})
for serviceName := range servicesRequiringUpdate {
var serviceMatch swarm.Service
for _, service := range services {
if service.Spec.Name == serviceName {
serviceMatch = service
break
}
}
if serviceMatch.ID == "" {
return fmt.Errorf("stopContainersAndRun: Couldn't find service with name %s", serviceName)
}
serviceMatch.Spec.TaskTemplate.ForceUpdate = 1
_, err := s.cli.ServiceUpdate(
s.ctx, serviceMatch.ID,
serviceMatch.Version, serviceMatch.Spec, types.ServiceUpdateOptions{},
2021-08-22 15:04:44 +02:00
)
if err != nil {
restartErrors = append(restartErrors, err)
}
}
}
if len(restartErrors) != 0 {
return fmt.Errorf(
"stopContainersAndRun: %d error(s) restarting containers and services: %w",
len(restartErrors),
err,
)
}
2021-08-23 07:07:44 +02:00
s.logger.Infof("Restarted %d container(s) and the matching service(s).", len(stoppedContainers))
return nil
}()
var stopErr error
if len(errors) != 0 {
stopErr = fmt.Errorf(
"stopContainersAndRun: %d errors stopping containers: %w",
len(errors),
err,
)
}
if stopErr != nil {
return stopErr
}
2021-08-21 21:26:27 +02:00
return thunk()
2021-08-21 19:26:42 +02:00
}
// takeBackup creates a tar archive of the configured backup location and
// saves it to disk.
2021-08-21 19:26:42 +02:00
func (s *script) takeBackup() error {
2021-08-22 22:02:19 +02:00
s.file = timeutil.Strftime(&s.start, s.file)
2021-08-22 19:37:48 +02:00
if err := targz.Compress(s.sources, s.file); err != nil {
2021-08-21 21:26:27 +02:00
return fmt.Errorf("takeBackup: error compressing backup folder: %w", err)
}
2021-08-23 07:07:44 +02:00
s.logger.Infof("Created backup of `%s` at `%s`.", s.sources, s.file)
2021-08-21 21:26:27 +02:00
return nil
2021-08-21 19:26:42 +02:00
}
// encryptBackup encrypts the backup file using PGP and the configured passphrase.
// In case no passphrase is given it returns early, leaving the backup file
// untouched.
2021-08-21 19:26:42 +02:00
func (s *script) encryptBackup() error {
2021-08-22 22:02:19 +02:00
if s.passphrase == nil {
2021-08-21 19:26:42 +02:00
return nil
}
2021-08-22 14:44:33 +02:00
buf := bytes.NewBuffer(nil)
_, name := path.Split(s.file)
2021-08-22 19:37:48 +02:00
pt, err := openpgp.SymmetricallyEncrypt(buf, []byte(s.passphrase), &openpgp.FileHints{
2021-08-22 14:44:33 +02:00
IsBinary: true,
FileName: name,
}, nil)
if err != nil {
return fmt.Errorf("encryptBackup: error encrypting backup file: %w", err)
}
unencrypted, err := ioutil.ReadFile(s.file)
if err != nil {
pt.Close()
return fmt.Errorf("encryptBackup: error reading unencrypted backup file: %w", err)
}
_, err = pt.Write(unencrypted)
if err != nil {
pt.Close()
return fmt.Errorf("encryptBackup: error writing backup contents: %w", err)
}
pt.Close()
gpgFile := fmt.Sprintf("%s.gpg", s.file)
if err := ioutil.WriteFile(gpgFile, buf.Bytes(), os.ModeAppend); err != nil {
return fmt.Errorf("encryptBackup: error writing encrypted version of backup: %w", err)
}
if err := os.Remove(s.file); err != nil {
return fmt.Errorf("encryptBackup: error removing unencrpyted backup: %w", err)
}
2021-08-22 22:02:19 +02:00
2021-08-22 14:44:33 +02:00
s.file = gpgFile
2021-08-23 07:07:44 +02:00
s.logger.Infof("Encrypted backup using given passphrase, saving as `%s`.", s.file)
2021-08-22 14:44:33 +02:00
return nil
2021-08-21 19:26:42 +02:00
}
// copyBackup makes sure the backup file is copied to both local and remote locations
// as per the given configuration.
2021-08-21 19:26:42 +02:00
func (s *script) copyBackup() error {
2021-08-21 21:26:27 +02:00
_, name := path.Split(s.file)
if s.bucket != "" {
_, err := s.mc.FPutObject(s.ctx, s.bucket, name, s.file, minio.PutObjectOptions{
2021-08-21 21:26:27 +02:00
ContentType: "application/tar+gzip",
})
if err != nil {
return fmt.Errorf("copyBackup: error uploading backup to remote storage: %w", err)
}
2021-08-23 07:07:44 +02:00
s.logger.Infof("Uploaded a copy of backup `%s` to bucket `%s`", s.file, s.bucket)
2021-08-21 21:26:27 +02:00
}
2021-08-22 15:04:44 +02:00
if _, err := os.Stat(s.archive); !os.IsNotExist(err) {
if err := copy(s.file, path.Join(s.archive, name)); err != nil {
return fmt.Errorf("copyBackup: error copying file to local archive: %w", err)
2021-08-21 21:26:27 +02:00
}
2021-08-23 07:07:44 +02:00
s.logger.Infof("Stored copy of backup `%s` in local archive `%s`", s.file, s.archive)
2021-08-21 21:26:27 +02:00
}
return nil
2021-08-21 19:26:42 +02:00
}
// cleanBackup removes the backup file from disk.
2021-08-21 19:26:42 +02:00
func (s *script) cleanBackup() error {
2021-08-21 21:26:27 +02:00
if err := os.Remove(s.file); err != nil {
return fmt.Errorf("cleanBackup: error removing file: %w", err)
}
2021-08-23 07:07:44 +02:00
s.logger.Info("Cleaned up local artifacts.")
2021-08-21 21:26:27 +02:00
return nil
2021-08-21 19:26:42 +02:00
}
// pruneOldBackups rotates away backups from local and remote storages using
// the given configuration. In case the given configuration would delete all
// backups, it does nothing instead.
func (s *script) pruneOldBackups() error {
2021-08-22 22:02:19 +02:00
if s.retentionDays == nil {
2021-08-21 19:26:42 +02:00
return nil
}
2021-08-22 22:02:19 +02:00
if s.leeway != nil {
s.logger.Infof("Sleeping for %s before pruning backups.", s.leeway)
time.Sleep(*s.leeway)
}
2021-08-22 22:02:19 +02:00
s.logger.Infof("Trying to prune backups older than %d day(s) now.", *s.retentionDays)
deadline := s.start.AddDate(0, 0, -*s.retentionDays)
2021-08-22 14:00:21 +02:00
if s.bucket != "" {
candidates := s.mc.ListObjects(s.ctx, s.bucket, minio.ListObjectsOptions{
2021-08-22 15:04:44 +02:00
WithMetadata: true,
2021-08-22 22:02:19 +02:00
Prefix: s.pruningPrefix,
2021-08-22 15:04:44 +02:00
})
var matches []minio.ObjectInfo
2021-08-22 16:41:06 +02:00
var lenCandidates int
2021-08-22 15:04:44 +02:00
for candidate := range candidates {
2021-08-22 16:41:06 +02:00
lenCandidates++
if candidate.Err != nil {
return fmt.Errorf("pruneOldBackups: error looking up candidates from remote storage: %w", candidate.Err)
}
2021-08-22 15:04:44 +02:00
if candidate.LastModified.Before(deadline) {
matches = append(matches, candidate)
}
}
2021-08-22 16:41:06 +02:00
if len(matches) != 0 && len(matches) != lenCandidates {
2021-08-22 15:04:44 +02:00
objectsCh := make(chan minio.ObjectInfo)
go func() {
2021-08-22 16:41:06 +02:00
for _, match := range matches {
objectsCh <- match
2021-08-22 15:04:44 +02:00
}
close(objectsCh)
2021-08-22 15:04:44 +02:00
}()
errChan := s.mc.RemoveObjects(s.ctx, s.bucket, objectsCh, minio.RemoveObjectsOptions{})
2021-08-22 15:04:44 +02:00
var errors []error
for result := range errChan {
if result.Err != nil {
errors = append(errors, result.Err)
}
}
if len(errors) != 0 {
return fmt.Errorf(
2021-08-22 22:02:19 +02:00
"pruneOldBackups: %d error(s) removing files from remote storage: %w",
2021-08-22 15:04:44 +02:00
len(errors),
errors[0],
)
}
2021-08-22 16:41:06 +02:00
s.logger.Infof(
2021-08-23 07:07:44 +02:00
"Pruned %d out of %d remote backup(s) as their age exceeded the configured retention period.",
2021-08-22 16:41:06 +02:00
len(matches),
lenCandidates,
)
} else if len(matches) != 0 && len(matches) == lenCandidates {
2021-08-22 19:37:48 +02:00
s.logger.Warnf(
"The current configuration would delete all %d remote backup copies. Refusing to do so, please check your configuration.",
len(matches),
)
2021-08-22 16:41:06 +02:00
} else {
2021-08-22 22:02:19 +02:00
s.logger.Infof("None of %d remote backup(s) were pruned.", lenCandidates)
2021-08-22 15:04:44 +02:00
}
}
2021-08-22 15:04:44 +02:00
if _, err := os.Stat(s.archive); !os.IsNotExist(err) {
2021-08-22 15:04:44 +02:00
candidates, err := filepath.Glob(
2021-08-22 22:02:19 +02:00
path.Join(s.archive, fmt.Sprintf("%s*", s.pruningPrefix)),
2021-08-22 14:00:21 +02:00
)
if err != nil {
2021-08-22 15:04:44 +02:00
return fmt.Errorf(
"pruneOldBackups: error looking up matching files, starting with: %w", err,
)
2021-08-22 14:00:21 +02:00
}
2021-08-22 15:04:44 +02:00
var matches []os.FileInfo
for _, candidate := range candidates {
fi, err := os.Stat(candidate)
2021-08-22 14:00:21 +02:00
if err != nil {
2021-08-22 15:04:44 +02:00
return fmt.Errorf(
"pruneOldBackups: error calling stat on file %s: %w",
candidate,
err,
)
2021-08-22 14:00:21 +02:00
}
if fi.ModTime().Before(deadline) {
2021-08-22 15:04:44 +02:00
matches = append(matches, fi)
2021-08-22 14:00:21 +02:00
}
}
2021-08-22 16:41:06 +02:00
if len(matches) != 0 && len(matches) != len(candidates) {
2021-08-22 15:04:44 +02:00
var errors []error
for _, candidate := range matches {
2021-08-22 14:00:21 +02:00
if err := os.Remove(candidate.Name()); err != nil {
2021-08-22 15:04:44 +02:00
errors = append(errors, err)
2021-08-22 14:00:21 +02:00
}
}
2021-08-22 15:04:44 +02:00
if len(errors) != 0 {
return fmt.Errorf(
2021-08-22 22:02:19 +02:00
"pruneOldBackups: %d error(s) deleting local files, starting with: %w",
2021-08-22 15:04:44 +02:00
len(errors),
errors[0],
)
}
2021-08-22 16:41:06 +02:00
s.logger.Infof(
2021-08-23 07:07:44 +02:00
"Pruned %d out of %d local backup(s) as their age exceeded the configured retention period.",
2021-08-22 16:41:06 +02:00
len(matches),
len(candidates),
)
} else if len(matches) != 0 && len(matches) == len(candidates) {
2021-08-22 19:37:48 +02:00
s.logger.Warnf(
"The current configuration would delete all %d local backup copies. Refusing to do so, please check your configuration.",
len(matches),
)
2021-08-22 16:41:06 +02:00
} else {
2021-08-22 22:02:19 +02:00
s.logger.Infof("None of %d local backup(s) were pruned.", len(candidates))
2021-08-22 14:00:21 +02:00
}
}
return nil
}
2021-08-22 19:37:48 +02:00
func (s *script) must(err error) {
if err != nil {
if s.logger == nil {
2021-08-21 21:26:27 +02:00
panic(err)
}
2021-08-22 19:37:48 +02:00
s.logger.Errorf("Fatal error running backup: %s", err)
os.Exit(1)
2021-08-21 21:26:27 +02:00
}
}
func copy(src, dst string) error {
in, err := os.Open(src)
if err != nil {
return err
}
defer in.Close()
out, err := os.Create(dst)
if err != nil {
return err
}
_, err = io.Copy(out, in)
if err != nil {
out.Close()
2021-08-21 21:26:27 +02:00
return err
}
return out.Close()
}