feat(backends/s3): add warmup support before repacks and restores (#5173)

* feat(backends/s3): add warmup support before repacks and restores

This commit introduces basic support for transitioning pack files stored
in cold storage to hot storage on S3 and S3-compatible providers.

To prevent unexpected behavior for existing users, the feature is gated
behind new flags:

- `s3.enable-restore`: opt-in flag (defaults to false)
- `s3.restore-days`: number of days for the restored objects to remain
  in hot storage (defaults to `7`)
- `s3.restore-timeout`: maximum time to wait for a single restoration
  (default to `1 day`)
- `s3.restore-tier`: retrieval tier at which the restore will be
  processed. (default to `Standard`)

As restoration times can be lengthy, this implementation preemptively
restores selected packs to prevent incessant restore-delays during
downloads. This is slightly sub-optimal as we could process packs
out-of-order (as soon as they're transitioned), but this would really
add too much complexity for a marginal gain in speed.

To maintain simplicity and prevent resources exhautions with lots of
packs, no new concurrency mechanisms or goroutines were added. This just
hooks gracefully into the existing routines.

**Limitations:**

- Tests against the backend were not written due to the lack of cold
  storage class support in MinIO. Testing was done manually on
  Scaleway's S3-compatible object storage. If necessary, we could
  explore testing with LocalStack or mocks, though this requires further
  discussion.
- Currently, this feature only warms up before restores and repacks
  (prune/copy), as those are the two main use-cases I came across.
  Support for other commands may be added in future iterations, as long
  as affected packs can be calculated in advance.
- The feature is gated behind a new alpha `s3-restore` feature flag to
  make it explicit that the feature is still wet behind the ears.
- There is no explicit user notification for ongoing pack restorations.
  While I think it is not necessary because of the opt-in flag, showing
  some notice may improve usability (but would probably require major
  refactoring in the progress bar which I didn't want to start). Another
  possibility would be to add a flag to send restores requests and fail
  early.

See https://github.com/restic/restic/issues/3202

* ui: warn user when files are warming up from cold storage

* refactor: remove the PacksWarmer struct

It's easier to handle multiple handles in the backend directly, and it
may open the door to reducing the number of requests made to the backend
in the future.
This commit is contained in:
Gilbert Gilb's 2025-02-01 19:26:27 +01:00 committed by GitHub
parent 9566e2db4a
commit 536ebefff4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
31 changed files with 680 additions and 123 deletions

View File

@ -0,0 +1,10 @@
Enhancement: Add warmup support on S3 backend before repacks and restores
Introduce S3 backend options for transitioning pack files from cold to hot
storage on S3 and S3-compatible providers. Note: only works before repacks
(prune/copy) and restore for now, and gated behind a new "s3-restore" feature
flag.
https://github.com/restic/restic/pull/5173
https://github.com/restic/restic/issues/3202
https://github.com/restic/restic/issues/2504

View File

@ -237,7 +237,15 @@ func copyTree(ctx context.Context, srcRepo restic.Repository, dstRepo restic.Rep
}
bar := newProgressMax(!quiet, uint64(len(packList)), "packs copied")
_, err = repository.Repack(ctx, srcRepo, dstRepo, packList, copyBlobs, bar)
_, err = repository.Repack(
ctx,
srcRepo,
dstRepo,
packList,
copyBlobs,
bar,
func(msg string, args ...interface{}) { fmt.Printf(msg+"\n", args...) },
)
bar.Done()
if err != nil {
return errors.Fatal(err.Error())

View File

@ -179,6 +179,12 @@ func runRestore(ctx context.Context, opts RestoreOptions, gopts GlobalOptions,
res.Warn = func(message string) {
msg.E("Warning: %s\n", message)
}
res.Info = func(message string) {
if gopts.JSON {
return
}
msg.P("Info: %s\n", message)
}
selectExcludeFilter := func(item string, isDir bool) (selectedForRestore bool, childMayBeSelected bool) {
matched := false

View File

@ -242,3 +242,33 @@ collect a list of all files, causing the following error:
List(data) returned error, retrying after 1s: [...]: request timeout
In this case you can increase the timeout using the ``--stuck-request-timeout`` option.
Are "cold storages" supported?
------------------------------
Generally, restic does not natively support "cold storage" solutions. However,
experimental support for restoring from **S3 Glacier** and **S3 Glacier Deep
Archive** storage classes is available:
.. code-block:: console
$ restic backup -o s3.storage-class=GLACIER somedir/
$ RESTIC_FEATURES=s3-restore restic restore -o s3.enable-restore=1 -o s3.restore-days=7 -o s3.restore-timeout=1d latest
**Notes:**
- This feature is still in early alpha stage. Expect arbitrary breaking changes
in the future (although we'll do our best-effort to avoid them).
- Expect restores to hang from 1 up to 42 hours depending on your storage
class, provider and luck. Restores from cold storages are known to be
time-consuming. You may need to adjust the `s3.restore-timeout` if a restore
operation takes more than 24 hours.
- Restic will prevent sending metadata files (such as config files, lock files
or tree blobs) to Glacier or Deep Archive. Standard class is used instead to
ensure normal and fast operations for most tasks.
- Currently, only the following commands are known to work:
- `backup`
- `copy`
- `prune`
- `restore`

View File

@ -475,3 +475,9 @@ func (be *Backend) Delete(ctx context.Context) error {
// Close does nothing
func (be *Backend) Close() error { return nil }
// Warmup not implemented
func (be *Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -335,3 +335,9 @@ func (be *b2Backend) Delete(ctx context.Context) error {
// Close does nothing
func (be *b2Backend) Close() error { return nil }
// Warmup not implemented
func (be *b2Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *b2Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -75,6 +75,21 @@ type Backend interface {
// Delete removes all data in the backend.
Delete(ctx context.Context) error
// Warmup ensures that the specified handles are ready for upcoming reads.
// This is particularly useful for transitioning files from cold to hot
// storage.
//
// The method is non-blocking. WarmupWait can be used to wait for
// completion.
//
// Returns:
// - Handles currently warming up.
// - An error if warmup fails.
Warmup(ctx context.Context, h []Handle) ([]Handle, error)
// WarmupWait waits until all given handles are warm.
WarmupWait(ctx context.Context, h []Handle) error
}
type Unwrapper interface {

View File

@ -258,3 +258,13 @@ func (b *Backend) List(ctx context.Context, t backend.FileType, fn func(f backen
return nil
}
// Warmup delegates to wrapped backend.
func (b *Backend) Warmup(ctx context.Context, h []backend.Handle) ([]backend.Handle, error) {
return b.Backend.Warmup(ctx, h)
}
// WarmupWait delegates to wrapped backend.
func (b *Backend) WarmupWait(ctx context.Context, h []backend.Handle) error {
return b.Backend.WarmupWait(ctx, h)
}

View File

@ -82,3 +82,9 @@ func (be *Backend) Load(ctx context.Context, h backend.Handle, length int, offse
func (be *Backend) Stat(ctx context.Context, h backend.Handle) (backend.FileInfo, error) {
return be.b.Stat(ctx, h)
}
// Warmup should not occur during dry-runs.
func (be *Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -363,3 +363,9 @@ func (be *Backend) Delete(ctx context.Context) error {
// Close does nothing.
func (be *Backend) Close() error { return nil }
// Warmup not implemented
func (be *Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -371,3 +371,9 @@ func (b *Local) Close() error {
// same function.
return nil
}
// Warmup not implemented
func (b *Local) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (b *Local) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -249,3 +249,9 @@ func (be *MemoryBackend) Delete(ctx context.Context) error {
func (be *MemoryBackend) Close() error {
return nil
}
// Warmup not implemented
func (be *MemoryBackend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *MemoryBackend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -20,6 +20,8 @@ type Backend struct {
ListFn func(ctx context.Context, t backend.FileType, fn func(backend.FileInfo) error) error
RemoveFn func(ctx context.Context, h backend.Handle) error
DeleteFn func(ctx context.Context) error
WarmupFn func(ctx context.Context, h []backend.Handle) ([]backend.Handle, error)
WarmupWaitFn func(ctx context.Context, h []backend.Handle) error
ConnectionsFn func() uint
HasherFn func() hash.Hash
HasAtomicReplaceFn func() bool
@ -150,5 +152,21 @@ func (m *Backend) Delete(ctx context.Context) error {
return m.DeleteFn(ctx)
}
func (m *Backend) Warmup(ctx context.Context, h []backend.Handle) ([]backend.Handle, error) {
if m.WarmupFn == nil {
return []backend.Handle{}, errors.New("not implemented")
}
return m.WarmupFn(ctx, h)
}
func (m *Backend) WarmupWait(ctx context.Context, h []backend.Handle) error {
if m.WarmupWaitFn == nil {
return errors.New("not implemented")
}
return m.WarmupWaitFn(ctx, h)
}
// Make sure that Backend implements the backend interface.
var _ backend.Backend = &Backend{}

View File

@ -340,3 +340,9 @@ func (be *Backend) Close() error {
debug.Log("wait for rclone returned: %v", be.waitResult)
return be.waitResult
}
// Warmup not implemented
func (be *Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -439,3 +439,9 @@ func (b *Backend) Close() error {
func (b *Backend) Delete(ctx context.Context) error {
return util.DefaultDelete(ctx, b)
}
// Warmup not implemented
func (b *Backend) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (b *Backend) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -289,3 +289,11 @@ func (be *Backend) List(ctx context.Context, t backend.FileType, fn func(backend
func (be *Backend) Unwrap() backend.Backend {
return be.Backend
}
// Warmup delegates to wrapped backend
func (be *Backend) Warmup(ctx context.Context, h []backend.Handle) ([]backend.Handle, error) {
return be.Backend.Warmup(ctx, h)
}
func (be *Backend) WarmupWait(ctx context.Context, h []backend.Handle) error {
return be.Backend.WarmupWait(ctx, h)
}

View File

@ -5,6 +5,7 @@ import (
"os"
"path"
"strings"
"time"
"github.com/restic/restic/internal/backend"
"github.com/restic/restic/internal/errors"
@ -23,6 +24,11 @@ type Config struct {
Layout string `option:"layout" help:"use this backend layout (default: auto-detect) (deprecated)"`
StorageClass string `option:"storage-class" help:"set S3 storage class (STANDARD, STANDARD_IA, ONEZONE_IA, INTELLIGENT_TIERING or REDUCED_REDUNDANCY)"`
EnableRestore bool `option:"enable-restore" help:"restore objects from GLACIER or DEEP_ARCHIVE storage classes (default: false, requires \"s3-restore\" feature flag)"`
RestoreDays int `option:"restore-days" help:"lifetime in days of restored object (default: 7)"`
RestoreTimeout time.Duration `option:"restore-timeout" help:"maximum time to wait for objects transition (default: 1d)"`
RestoreTier string `option:"restore-tier" help:"Retrieval tier at which the restore will be processed. (Standard, Bulk or Expedited) (default: Standard)"`
Connections uint `option:"connections" help:"set a limit for the number of concurrent connections (default: 5)"`
MaxRetries uint `option:"retries" help:"set the number of retries attempted"`
Region string `option:"region" help:"set region"`
@ -34,8 +40,12 @@ type Config struct {
// NewConfig returns a new Config with the default values filled in.
func NewConfig() Config {
return Config{
Connections: 5,
ListObjectsV1: false,
Connections: 5,
ListObjectsV1: false,
EnableRestore: false,
RestoreDays: 7,
RestoreTimeout: 24 * time.Hour,
RestoreTier: "Standard",
}
}

View File

@ -3,117 +3,117 @@ package s3
import (
"strings"
"testing"
"time"
"github.com/restic/restic/internal/backend/test"
)
func newTestConfig(cfg Config) Config {
if cfg.Connections == 0 {
cfg.Connections = 5
}
if cfg.RestoreDays == 0 {
cfg.RestoreDays = 7
}
if cfg.RestoreTimeout == 0 {
cfg.RestoreTimeout = 24 * time.Hour
}
if cfg.RestoreTier == "" {
cfg.RestoreTier = "Standard"
}
return cfg
}
var configTests = []test.ConfigTestData[Config]{
{S: "s3://eu-central-1/bucketname", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "",
Connections: 5,
}},
{S: "s3://eu-central-1/bucketname/", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "",
Connections: 5,
}},
{S: "s3://eu-central-1/bucketname/prefix/directory", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "prefix/directory",
Connections: 5,
}},
{S: "s3://eu-central-1/bucketname/prefix/directory/", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "prefix/directory",
Connections: 5,
}},
{S: "s3:eu-central-1/foobar", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:eu-central-1/foobar/", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:eu-central-1/foobar/prefix/directory", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "prefix/directory",
Connections: 5,
}},
{S: "s3:eu-central-1/foobar/prefix/directory/", Cfg: Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "prefix/directory",
Connections: 5,
}},
{S: "s3:hostname.foo/foobar", Cfg: Config{
Endpoint: "hostname.foo",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:hostname.foo/foobar/prefix/directory", Cfg: Config{
Endpoint: "hostname.foo",
Bucket: "foobar",
Prefix: "prefix/directory",
Connections: 5,
}},
{S: "s3:https://hostname/foobar", Cfg: Config{
Endpoint: "hostname",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:https://hostname:9999/foobar", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:https://hostname:9999/foobar/", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
Connections: 5,
}},
{S: "s3:http://hostname:9999/foobar", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
UseHTTP: true,
Connections: 5,
}},
{S: "s3:http://hostname:9999/foobar/", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
UseHTTP: true,
Connections: 5,
}},
{S: "s3:http://hostname:9999/bucket/prefix/directory", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "bucket",
Prefix: "prefix/directory",
UseHTTP: true,
Connections: 5,
}},
{S: "s3:http://hostname:9999/bucket/prefix/directory/", Cfg: Config{
Endpoint: "hostname:9999",
Bucket: "bucket",
Prefix: "prefix/directory",
UseHTTP: true,
Connections: 5,
}},
{S: "s3://eu-central-1/bucketname", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "",
})},
{S: "s3://eu-central-1/bucketname/", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "",
})},
{S: "s3://eu-central-1/bucketname/prefix/directory", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "prefix/directory",
})},
{S: "s3://eu-central-1/bucketname/prefix/directory/", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "bucketname",
Prefix: "prefix/directory",
})},
{S: "s3:eu-central-1/foobar", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:eu-central-1/foobar/", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:eu-central-1/foobar/prefix/directory", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "prefix/directory",
})},
{S: "s3:eu-central-1/foobar/prefix/directory/", Cfg: newTestConfig(Config{
Endpoint: "eu-central-1",
Bucket: "foobar",
Prefix: "prefix/directory",
})},
{S: "s3:hostname.foo/foobar", Cfg: newTestConfig(Config{
Endpoint: "hostname.foo",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:hostname.foo/foobar/prefix/directory", Cfg: newTestConfig(Config{
Endpoint: "hostname.foo",
Bucket: "foobar",
Prefix: "prefix/directory",
})},
{S: "s3:https://hostname/foobar", Cfg: newTestConfig(Config{
Endpoint: "hostname",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:https://hostname:9999/foobar", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:https://hostname:9999/foobar/", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
})},
{S: "s3:http://hostname:9999/foobar", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
UseHTTP: true,
})},
{S: "s3:http://hostname:9999/foobar/", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "foobar",
Prefix: "",
UseHTTP: true,
})},
{S: "s3:http://hostname:9999/bucket/prefix/directory", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "bucket",
Prefix: "prefix/directory",
UseHTTP: true,
})},
{S: "s3:http://hostname:9999/bucket/prefix/directory/", Cfg: newTestConfig(Config{
Endpoint: "hostname:9999",
Bucket: "bucket",
Prefix: "prefix/directory",
UseHTTP: true,
})},
}
func TestParseConfig(t *testing.T) {

View File

@ -8,8 +8,11 @@ import (
"net/http"
"os"
"path"
"slices"
"strings"
"time"
"github.com/cenkalti/backoff/v4"
"github.com/restic/restic/internal/backend"
"github.com/restic/restic/internal/backend/layout"
"github.com/restic/restic/internal/backend/location"
@ -32,6 +35,17 @@ type Backend struct {
// make sure that *Backend implements backend.Backend
var _ backend.Backend = &Backend{}
var archiveClasses = []string{"GLACIER", "DEEP_ARCHIVE"}
type warmupStatus int
const (
warmupStatusCold warmupStatus = iota
warmupStatusWarmingUp
warmupStatusWarm
warmupStatusLukewarm
)
func NewFactory() location.Factory {
return location.NewHTTPBackendFactory("s3", ParseConfig, location.NoPassword, Create, Open)
}
@ -39,6 +53,10 @@ func NewFactory() location.Factory {
func open(cfg Config, rt http.RoundTripper) (*Backend, error) {
debug.Log("open, config %#v", cfg)
if cfg.EnableRestore && !feature.Flag.Enabled(feature.S3Restore) {
return nil, fmt.Errorf("feature flag `s3-restore` is required to use `-o s3.enable-restore=true`")
}
if cfg.KeyID == "" && cfg.Secret.String() != "" {
return nil, errors.Fatalf("unable to open S3 backend: Key ID ($AWS_ACCESS_KEY_ID) is empty")
} else if cfg.KeyID != "" && cfg.Secret.String() == "" {
@ -266,9 +284,9 @@ func (be *Backend) Path() string {
// For archive storage classes, only data files are stored using that class; metadata
// must remain instantly accessible.
func (be *Backend) useStorageClass(h backend.Handle) bool {
notArchiveClass := be.cfg.StorageClass != "GLACIER" && be.cfg.StorageClass != "DEEP_ARCHIVE"
isDataFile := h.Type == backend.PackFile && !h.IsMetadata
return isDataFile || notArchiveClass
isArchiveClass := slices.Contains(archiveClasses, be.cfg.StorageClass)
return !isArchiveClass || isDataFile
}
// Save stores data in the backend at the handle.
@ -440,3 +458,148 @@ func (be *Backend) Delete(ctx context.Context) error {
// Close does nothing
func (be *Backend) Close() error { return nil }
// Warmup transitions handles from cold to hot storage if needed.
func (be *Backend) Warmup(ctx context.Context, handles []backend.Handle) ([]backend.Handle, error) {
handlesWarmingUp := []backend.Handle{}
if be.cfg.EnableRestore {
for _, h := range handles {
filename := be.Filename(h)
isWarmingUp, err := be.requestRestore(ctx, filename)
if err != nil {
return handlesWarmingUp, err
}
if isWarmingUp {
debug.Log("s3 file is being restored: %s", filename)
handlesWarmingUp = append(handlesWarmingUp, h)
}
}
}
return handlesWarmingUp, nil
}
// requestRestore sends a glacier restore request on a given file.
func (be *Backend) requestRestore(ctx context.Context, filename string) (bool, error) {
objectInfo, err := be.client.StatObject(ctx, be.cfg.Bucket, filename, minio.StatObjectOptions{})
if err != nil {
return false, err
}
ws := be.getWarmupStatus(objectInfo)
switch ws {
case warmupStatusWarm:
return false, nil
case warmupStatusWarmingUp:
return true, nil
}
opts := minio.RestoreRequest{}
opts.SetDays(be.cfg.RestoreDays)
opts.SetGlacierJobParameters(minio.GlacierJobParameters{Tier: minio.TierType(be.cfg.RestoreTier)})
if err := be.client.RestoreObject(ctx, be.cfg.Bucket, filename, "", opts); err != nil {
var e minio.ErrorResponse
if errors.As(err, &e) {
switch e.Code {
case "InvalidObjectState":
return false, nil
case "RestoreAlreadyInProgress":
return true, nil
}
}
return false, err
}
isWarmingUp := ws != warmupStatusLukewarm
return isWarmingUp, nil
}
// getWarmupStatus returns the warmup status of the provided object.
func (be *Backend) getWarmupStatus(objectInfo minio.ObjectInfo) warmupStatus {
// We can't use objectInfo.StorageClass to get the storage class of the
// object because this field is only set during ListObjects operations.
// The response header is the documented way to get the storage class
// for GetObject/StatObject operations.
storageClass := objectInfo.Metadata.Get("X-Amz-Storage-Class")
isArchiveClass := slices.Contains(archiveClasses, storageClass)
if !isArchiveClass {
return warmupStatusWarm
}
restore := objectInfo.Restore
if restore != nil {
if restore.OngoingRestore {
return warmupStatusWarmingUp
}
minExpiryTime := time.Now().Add(time.Duration(be.cfg.RestoreDays) * 24 * time.Hour)
expiryTime := restore.ExpiryTime
if !expiryTime.IsZero() {
if minExpiryTime.Before(expiryTime) {
return warmupStatusWarm
}
return warmupStatusLukewarm
}
}
return warmupStatusCold
}
// WarmupWait waits until all handles are in hot storage.
func (be *Backend) WarmupWait(ctx context.Context, handles []backend.Handle) error {
timeoutCtx, timeoutCtxCancel := context.WithTimeout(ctx, be.cfg.RestoreTimeout)
defer timeoutCtxCancel()
if be.cfg.EnableRestore {
for _, h := range handles {
filename := be.Filename(h)
err := be.waitForRestore(timeoutCtx, filename)
if err != nil {
return err
}
debug.Log("s3 file is restored: %s", filename)
}
}
return nil
}
// waitForRestore waits for a given file to be restored.
func (be *Backend) waitForRestore(ctx context.Context, filename string) error {
for {
var objectInfo minio.ObjectInfo
// Restore requests can last many hours, therefore network may fail
// temporarily. We don't need to die in such even.
b := backoff.WithMaxRetries(backoff.NewExponentialBackOff(), 10)
b = backoff.WithContext(b, ctx)
err := backoff.Retry(
func() (err error) {
objectInfo, err = be.client.StatObject(ctx, be.cfg.Bucket, filename, minio.StatObjectOptions{})
return
},
b,
)
if err != nil {
return err
}
ws := be.getWarmupStatus(objectInfo)
switch ws {
case warmupStatusLukewarm:
fallthrough
case warmupStatusWarm:
return nil
case warmupStatusCold:
return errors.New("waiting on S3 handle that is not warming up")
}
select {
case <-time.After(1 * time.Minute):
case <-ctx.Done():
return ctx.Err()
}
}
}

View File

@ -588,3 +588,9 @@ func (r *SFTP) deleteRecursive(ctx context.Context, name string) error {
func (r *SFTP) Delete(ctx context.Context) error {
return r.deleteRecursive(ctx, r.p)
}
// Warmup not implemented
func (r *SFTP) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (r *SFTP) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -269,3 +269,9 @@ func (be *beSwift) Delete(ctx context.Context) error {
// Close does nothing
func (be *beSwift) Close() error { return nil }
// Warmup not implemented
func (be *beSwift) Warmup(_ context.Context, _ []backend.Handle) ([]backend.Handle, error) {
return []backend.Handle{}, nil
}
func (be *beSwift) WarmupWait(_ context.Context, _ []backend.Handle) error { return nil }

View File

@ -9,6 +9,7 @@ const (
DeviceIDForHardlinks FlagName = "device-id-for-hardlinks"
ExplicitS3AnonymousAuth FlagName = "explicit-s3-anonymous-auth"
SafeForgetKeepTags FlagName = "safe-forget-keep-tags"
S3Restore FlagName = "s3-restore"
)
func init() {
@ -17,5 +18,6 @@ func init() {
DeviceIDForHardlinks: {Type: Alpha, Description: "store deviceID only for hardlinks to reduce metadata changes for example when using btrfs subvolumes. Will be removed in a future restic version after repository format 3 is available"},
ExplicitS3AnonymousAuth: {Type: Stable, Description: "forbid anonymous S3 authentication unless `-o s3.unsafe-anonymous-auth=true` is set"},
SafeForgetKeepTags: {Type: Stable, Description: "prevent deleting all snapshots if the tag passed to `forget --keep-tags tagname` does not exist"},
S3Restore: {Type: Alpha, Description: "restore S3 objects from cold storage classes when `-o s3.enable-restore=true` is set"},
})
}

View File

@ -557,7 +557,7 @@ func (plan *PrunePlan) Execute(ctx context.Context, printer progress.Printer) er
printer.P("repacking packs\n")
bar := printer.NewCounter("packs repacked")
bar.SetMax(uint64(len(plan.repackPacks)))
_, err := Repack(ctx, repo, repo, plan.repackPacks, plan.keepBlobs, bar)
_, err := Repack(ctx, repo, repo, plan.repackPacks, plan.keepBlobs, bar, printer.P)
bar.Done()
if err != nil {
return errors.Fatal(err.Error())

View File

@ -6,6 +6,7 @@ import (
"github.com/restic/restic/internal/debug"
"github.com/restic/restic/internal/errors"
"github.com/restic/restic/internal/feature"
"github.com/restic/restic/internal/restic"
"github.com/restic/restic/internal/ui/progress"
@ -18,6 +19,8 @@ type repackBlobSet interface {
Len() int
}
type LogFunc func(msg string, args ...interface{})
// Repack takes a list of packs together with a list of blobs contained in
// these packs. Each pack is loaded and the blobs listed in keepBlobs is saved
// into a new pack. Returned is the list of obsolete packs which can then
@ -25,9 +28,21 @@ type repackBlobSet interface {
//
// The map keepBlobs is modified by Repack, it is used to keep track of which
// blobs have been processed.
func Repack(ctx context.Context, repo restic.Repository, dstRepo restic.Repository, packs restic.IDSet, keepBlobs repackBlobSet, p *progress.Counter) (obsoletePacks restic.IDSet, err error) {
func Repack(
ctx context.Context,
repo restic.Repository,
dstRepo restic.Repository,
packs restic.IDSet,
keepBlobs repackBlobSet,
p *progress.Counter,
logf LogFunc,
) (obsoletePacks restic.IDSet, err error) {
debug.Log("repacking %d packs while keeping %d blobs", len(packs), keepBlobs.Len())
if logf == nil {
logf = func(_ string, _ ...interface{}) {}
}
if repo == dstRepo && dstRepo.Connections() < 2 {
return nil, errors.New("repack step requires a backend connection limit of at least two")
}
@ -37,7 +52,7 @@ func Repack(ctx context.Context, repo restic.Repository, dstRepo restic.Reposito
dstRepo.StartPackUploader(wgCtx, wg)
wg.Go(func() error {
var err error
obsoletePacks, err = repack(wgCtx, repo, dstRepo, packs, keepBlobs, p)
obsoletePacks, err = repack(wgCtx, repo, dstRepo, packs, keepBlobs, p, logf)
return err
})
@ -47,9 +62,30 @@ func Repack(ctx context.Context, repo restic.Repository, dstRepo restic.Reposito
return obsoletePacks, nil
}
func repack(ctx context.Context, repo restic.Repository, dstRepo restic.Repository, packs restic.IDSet, keepBlobs repackBlobSet, p *progress.Counter) (obsoletePacks restic.IDSet, err error) {
func repack(
ctx context.Context,
repo restic.Repository,
dstRepo restic.Repository,
packs restic.IDSet,
keepBlobs repackBlobSet,
p *progress.Counter,
logf LogFunc,
) (obsoletePacks restic.IDSet, err error) {
wg, wgCtx := errgroup.WithContext(ctx)
if feature.Flag.Enabled(feature.S3Restore) {
job, err := repo.StartWarmup(ctx, packs)
if err != nil {
return nil, err
}
if job.HandleCount() != 0 {
logf("warming up %d packs from cold storage, this may take a while...", job.HandleCount())
if err := job.Wait(ctx); err != nil {
return nil, err
}
}
}
var keepMutex sync.Mutex
downloadQueue := make(chan restic.PackBlobs)
wg.Go(func() error {

View File

@ -160,7 +160,7 @@ func findPacksForBlobs(t *testing.T, repo restic.Repository, blobs restic.BlobSe
}
func repack(t *testing.T, repo restic.Repository, be backend.Backend, packs restic.IDSet, blobs restic.BlobSet) {
repackedBlobs, err := repository.Repack(context.TODO(), repo, repo, packs, blobs, nil)
repackedBlobs, err := repository.Repack(context.TODO(), repo, repo, packs, blobs, nil, nil)
if err != nil {
t.Fatal(err)
}
@ -279,7 +279,7 @@ func testRepackCopy(t *testing.T, version uint) {
_, keepBlobs := selectBlobs(t, random, repo, 0.2)
copyPacks := findPacksForBlobs(t, repo, keepBlobs)
_, err := repository.Repack(context.TODO(), repoWrapped, dstRepoWrapped, copyPacks, keepBlobs, nil)
_, err := repository.Repack(context.TODO(), repoWrapped, dstRepoWrapped, copyPacks, keepBlobs, nil, nil)
if err != nil {
t.Fatal(err)
}
@ -318,7 +318,7 @@ func testRepackWrongBlob(t *testing.T, version uint) {
_, keepBlobs := selectBlobs(t, random, repo, 0)
rewritePacks := findPacksForBlobs(t, repo, keepBlobs)
_, err := repository.Repack(context.TODO(), repo, repo, rewritePacks, keepBlobs, nil)
_, err := repository.Repack(context.TODO(), repo, repo, rewritePacks, keepBlobs, nil, nil)
if err == nil {
t.Fatal("expected repack to fail but got no error")
}
@ -366,7 +366,7 @@ func testRepackBlobFallback(t *testing.T, version uint) {
rtest.OK(t, repo.Flush(context.Background()))
// repack must fallback to valid copy
_, err = repository.Repack(context.TODO(), repo, repo, rewritePacks, keepBlobs, nil)
_, err = repository.Repack(context.TODO(), repo, repo, rewritePacks, keepBlobs, nil, nil)
rtest.OK(t, err)
keepBlobs = restic.NewBlobSet(restic.BlobHandle{Type: restic.DataBlob, ID: id})

View File

@ -0,0 +1,39 @@
package repository
import (
"context"
"github.com/restic/restic/internal/backend"
"github.com/restic/restic/internal/restic"
)
type WarmupJob struct {
repo *Repository
handlesWarmingUp []backend.Handle
}
// HandleCount returns the number of handles that are currently warming up.
func (job *WarmupJob) HandleCount() int {
return len(job.handlesWarmingUp)
}
// Wait waits for all handles to be warm.
func (job *WarmupJob) Wait(ctx context.Context) error {
return job.repo.be.WarmupWait(ctx, job.handlesWarmingUp)
}
// StartWarmup creates a new warmup job, requesting the backend to warmup the specified packs.
func (repo *Repository) StartWarmup(ctx context.Context, packs restic.IDSet) (restic.WarmupJob, error) {
handles := make([]backend.Handle, 0, len(packs))
for pack := range packs {
handles = append(
handles,
backend.Handle{Type: restic.PackFile, Name: pack.String()},
)
}
handlesWarmingUp, err := repo.be.Warmup(ctx, handles)
return &WarmupJob{
repo: repo,
handlesWarmingUp: handlesWarmingUp,
}, err
}

View File

@ -0,0 +1,73 @@
package repository
import (
"context"
"testing"
"github.com/restic/restic/internal/backend"
"github.com/restic/restic/internal/backend/mock"
"github.com/restic/restic/internal/restic"
)
func TestWarmupRepository(t *testing.T) {
warmupCalls := [][]backend.Handle{}
warmupWaitCalls := [][]backend.Handle{}
simulateWarmingUp := false
be := mock.NewBackend()
be.WarmupFn = func(ctx context.Context, handles []backend.Handle) ([]backend.Handle, error) {
warmupCalls = append(warmupCalls, handles)
if simulateWarmingUp {
return handles, nil
}
return []backend.Handle{}, nil
}
be.WarmupWaitFn = func(ctx context.Context, handles []backend.Handle) error {
warmupWaitCalls = append(warmupWaitCalls, handles)
return nil
}
repo, _ := New(be, Options{})
id1, _ := restic.ParseID("1111111111111111111111111111111111111111111111111111111111111111")
id2, _ := restic.ParseID("2222222222222222222222222222222222222222222222222222222222222222")
id3, _ := restic.ParseID("3333333333333333333333333333333333333333333333333333333333333333")
job, err := repo.StartWarmup(context.TODO(), restic.NewIDSet(id1, id2))
if err != nil {
t.Fatalf("error when starting warmup: %v", err)
}
if len(warmupCalls) != 1 {
t.Fatalf("expected %d calls to warmup, got %d", 1, len(warmupCalls))
}
if len(warmupCalls[0]) != 2 {
t.Fatalf("expected warmup on %d handles, got %d", 2, len(warmupCalls[0]))
}
if job.HandleCount() != 0 {
t.Fatalf("expected all files to be warm, got %d cold", job.HandleCount())
}
simulateWarmingUp = true
job, err = repo.StartWarmup(context.TODO(), restic.NewIDSet(id3))
if err != nil {
t.Fatalf("error when starting warmup: %v", err)
}
if len(warmupCalls) != 2 {
t.Fatalf("expected %d calls to warmup, got %d", 2, len(warmupCalls))
}
if len(warmupCalls[1]) != 1 {
t.Fatalf("expected warmup on %d handles, got %d", 1, len(warmupCalls[1]))
}
if job.HandleCount() != 1 {
t.Fatalf("expected %d file to be warming up, got %d", 1, job.HandleCount())
}
if err := job.Wait(context.TODO()); err != nil {
t.Fatalf("error when waiting warmup: %v", err)
}
if len(warmupWaitCalls) != 1 {
t.Fatalf("expected %d calls to warmupWait, got %d", 1, len(warmupCalls))
}
if len(warmupWaitCalls[0]) != 1 {
t.Fatalf("expected warmupWait to be called with %d handles, got %d", 1, len(warmupWaitCalls[0]))
}
}

View File

@ -60,6 +60,9 @@ type Repository interface {
SaveUnpacked(ctx context.Context, t WriteableFileType, buf []byte) (ID, error)
// RemoveUnpacked removes a file from the repository. This will eventually be restricted to deleting only snapshots.
RemoveUnpacked(ctx context.Context, t WriteableFileType, id ID) error
// StartWarmup creates a new warmup job, requesting the backend to warmup the specified packs.
StartWarmup(ctx context.Context, packs IDSet) (WarmupJob, error)
}
type FileType = backend.FileType
@ -157,3 +160,10 @@ type Unpacked[FT FileTypes] interface {
type ListBlobser interface {
ListBlobs(ctx context.Context, fn func(PackedBlob)) error
}
type WarmupJob interface {
// HandleCount returns the number of handles that are currently warming up.
HandleCount() int
// Wait waits for all handles to be warm.
Wait(ctx context.Context) error
}

View File

@ -2,6 +2,7 @@ package restorer
import (
"context"
"fmt"
"path/filepath"
"sync"
@ -9,6 +10,7 @@ import (
"github.com/restic/restic/internal/debug"
"github.com/restic/restic/internal/errors"
"github.com/restic/restic/internal/feature"
"github.com/restic/restic/internal/repository"
"github.com/restic/restic/internal/restic"
"github.com/restic/restic/internal/ui/restore"
@ -41,12 +43,15 @@ type packInfo struct {
}
type blobsLoaderFn func(ctx context.Context, packID restic.ID, blobs []restic.Blob, handleBlobFn func(blob restic.BlobHandle, buf []byte, err error) error) error
type startWarmupFn func(context.Context, restic.IDSet) (restic.WarmupJob, error)
// fileRestorer restores set of files
type fileRestorer struct {
idx func(restic.BlobType, restic.ID) []restic.PackedBlob
blobsLoader blobsLoaderFn
startWarmup startWarmupFn
workerCount int
filesWriter *filesWriter
zeroChunk restic.ID
@ -58,6 +63,7 @@ type fileRestorer struct {
dst string
files []*fileInfo
Error func(string, error) error
Info func(string)
}
func newFileRestorer(dst string,
@ -66,6 +72,7 @@ func newFileRestorer(dst string,
connections uint,
sparse bool,
allowRecursiveDelete bool,
startWarmup startWarmupFn,
progress *restore.Progress) *fileRestorer {
// as packs are streamed the concurrency is limited by IO
@ -74,6 +81,7 @@ func newFileRestorer(dst string,
return &fileRestorer{
idx: idx,
blobsLoader: blobsLoader,
startWarmup: startWarmup,
filesWriter: newFilesWriter(workerCount, allowRecursiveDelete),
zeroChunk: repository.ZeroChunk(),
sparse: sparse,
@ -82,6 +90,7 @@ func newFileRestorer(dst string,
workerCount: workerCount,
dst: dst,
Error: restorerAbortOnAllErrors,
Info: func(_ string) {},
}
}
@ -192,6 +201,19 @@ func (r *fileRestorer) restoreFiles(ctx context.Context) error {
// drop no longer necessary file list
r.files = nil
if feature.Flag.Enabled(feature.S3Restore) {
warmupJob, err := r.startWarmup(ctx, restic.NewIDSet(packOrder...))
if err != nil {
return err
}
if warmupJob.HandleCount() != 0 {
r.Info(fmt.Sprintf("warming up %d packs from cold storage, this may take a while...", warmupJob.HandleCount()))
if err := warmupJob.Wait(ctx); err != nil {
return err
}
}
}
wg, ctx := errgroup.WithContext(ctx)
downloadCh := make(chan *packInfo)

View File

@ -9,6 +9,7 @@ import (
"testing"
"github.com/restic/restic/internal/errors"
"github.com/restic/restic/internal/feature"
"github.com/restic/restic/internal/restic"
rtest "github.com/restic/restic/internal/test"
)
@ -23,6 +24,11 @@ type TestFile struct {
blobs []TestBlob
}
type TestWarmupJob struct {
handlesCount int
waitCalled bool
}
type TestRepo struct {
packsIDToData map[restic.ID][]byte
@ -31,6 +37,8 @@ type TestRepo struct {
files []*fileInfo
filesPathToContent map[string]string
warmupJobs []*TestWarmupJob
//
loader blobsLoaderFn
}
@ -44,6 +52,21 @@ func (i *TestRepo) fileContent(file *fileInfo) string {
return i.filesPathToContent[file.location]
}
func (i *TestRepo) StartWarmup(ctx context.Context, packs restic.IDSet) (restic.WarmupJob, error) {
job := TestWarmupJob{handlesCount: len(packs)}
i.warmupJobs = append(i.warmupJobs, &job)
return &job, nil
}
func (job *TestWarmupJob) HandleCount() int {
return job.handlesCount
}
func (job *TestWarmupJob) Wait(_ context.Context) error {
job.waitCalled = true
return nil
}
func newTestRepo(content []TestFile) *TestRepo {
type Pack struct {
name string
@ -111,6 +134,7 @@ func newTestRepo(content []TestFile) *TestRepo {
blobs: blobs,
files: files,
filesPathToContent: filesPathToContent,
warmupJobs: []*TestWarmupJob{},
}
repo.loader = func(ctx context.Context, packID restic.ID, blobs []restic.Blob, handleBlobFn func(blob restic.BlobHandle, buf []byte, err error) error) error {
blobs = append([]restic.Blob{}, blobs...)
@ -141,10 +165,12 @@ func newTestRepo(content []TestFile) *TestRepo {
}
func restoreAndVerify(t *testing.T, tempdir string, content []TestFile, files map[string]bool, sparse bool) {
defer feature.TestSetFlag(t, feature.Flag, feature.S3Restore, true)()
t.Helper()
repo := newTestRepo(content)
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, sparse, false, nil)
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, sparse, false, repo.StartWarmup, nil)
if files == nil {
r.files = repo.files
@ -177,6 +203,15 @@ func verifyRestore(t *testing.T, r *fileRestorer, repo *TestRepo) {
t.Errorf("file %v has wrong content: want %q, got %q", file.location, content, data)
}
}
if len(repo.warmupJobs) == 0 {
t.Errorf("warmup did not occur")
}
for i, warmupJob := range repo.warmupJobs {
if !warmupJob.waitCalled {
t.Errorf("warmup job %d was not waited", i)
}
}
}
func TestFileRestorerBasic(t *testing.T) {
@ -285,7 +320,7 @@ func TestErrorRestoreFiles(t *testing.T) {
return loadError
}
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, false, false, nil)
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, false, false, repo.StartWarmup, nil)
r.files = repo.files
err := r.restoreFiles(context.TODO())
@ -326,7 +361,7 @@ func TestFatalDownloadError(t *testing.T) {
})
}
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, false, false, nil)
r := newFileRestorer(tempdir, repo.loader, repo.Lookup, 2, false, false, repo.StartWarmup, nil)
r.files = repo.files
var errors []string

View File

@ -28,6 +28,7 @@ type Restorer struct {
Error func(location string, err error) error
Warn func(message string)
Info func(message string)
// SelectFilter determines whether the item is selectedForRestore or whether a childMayBeSelected.
// selectedForRestore must not depend on isDir as `removeUnexpectedFiles` always passes false to isDir.
SelectFilter func(item string, isDir bool) (selectedForRestore bool, childMayBeSelected bool)
@ -357,8 +358,9 @@ func (res *Restorer) RestoreTo(ctx context.Context, dst string) (uint64, error)
idx := NewHardlinkIndex[string]()
filerestorer := newFileRestorer(dst, res.repo.LoadBlobsFromPack, res.repo.LookupBlob,
res.repo.Connections(), res.opts.Sparse, res.opts.Delete, res.opts.Progress)
res.repo.Connections(), res.opts.Sparse, res.opts.Delete, res.repo.StartWarmup, res.opts.Progress)
filerestorer.Error = res.Error
filerestorer.Info = res.Info
debug.Log("first pass for %q", dst)