storj/satellite/metabase/commit.go
Michal Niewrzal 4d9c9138ce satellite/metainfo: use multiple object versions internally
With this change we are switching methods to begin object, from
BeginObjectExactVersion to BeginObjectNextVersion. Main implication
is that from now it will be possible to have object with version
different than 1. New object will always get first available version.

Main reason to do this it to avoid deleting existing object during
reuploading object. Now we can create multiple pending objects but
only last committed will be available to the user. Any previous
committed object will be deleted.Because of that we moved logic to
delete existing object from BeginObject to CommitoObject request.

New logic is behind feature flat to be able to test it well first
before enablng on production.

Fixes https://github.com/storj/storj/issues/4871

Change-Id: I2dd9c7364fd93796a05ef607bda9c39a741e6a89
2022-10-06 15:19:02 +00:00

725 lines
22 KiB
Go

// Copyright (C) 2020 Storj Labs, Inc.
// See LICENSE for copying information.
package metabase
import (
"context"
"database/sql"
"errors"
"time"
pgxerrcode "github.com/jackc/pgerrcode"
"github.com/zeebo/errs"
"go.uber.org/zap"
"storj.io/common/memory"
"storj.io/common/storj"
"storj.io/private/dbutil/pgutil/pgerrcode"
"storj.io/private/dbutil/txutil"
"storj.io/private/tagsql"
)
// we need to disable PlainSize validation for old uplinks.
const validatePlainSize = false
const defaultZombieDeletionPeriod = 24 * time.Hour
var (
// ErrInvalidRequest is used to indicate invalid requests.
ErrInvalidRequest = errs.Class("metabase: invalid request")
// ErrConflict is used to indicate conflict with the request.
ErrConflict = errs.Class("metabase: conflict")
)
// BeginObjectNextVersion contains arguments necessary for starting an object upload.
type BeginObjectNextVersion struct {
ObjectStream
ExpiresAt *time.Time
ZombieDeletionDeadline *time.Time
EncryptedMetadata []byte // optional
EncryptedMetadataNonce []byte // optional
EncryptedMetadataEncryptedKey []byte // optional
Encryption storj.EncryptionParameters
}
// Verify verifies get object request fields.
func (opts *BeginObjectNextVersion) Verify() error {
if err := opts.ObjectStream.Verify(); err != nil {
return err
}
if opts.Version != NextVersion {
return ErrInvalidRequest.New("Version should be metabase.NextVersion")
}
if opts.EncryptedMetadata == nil && (opts.EncryptedMetadataNonce != nil || opts.EncryptedMetadataEncryptedKey != nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be not set if EncryptedMetadata is not set")
} else if opts.EncryptedMetadata != nil && (opts.EncryptedMetadataNonce == nil || opts.EncryptedMetadataEncryptedKey == nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be set if EncryptedMetadata is set")
}
return nil
}
// BeginObjectNextVersion adds a pending object to the database, with automatically assigned version.
func (db *DB) BeginObjectNextVersion(ctx context.Context, opts BeginObjectNextVersion) (object Object, err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.Verify(); err != nil {
return Object{}, err
}
if opts.ZombieDeletionDeadline == nil {
deadline := time.Now().Add(defaultZombieDeletionPeriod)
opts.ZombieDeletionDeadline = &deadline
}
object = Object{
ObjectStream: ObjectStream{
ProjectID: opts.ProjectID,
BucketName: opts.BucketName,
ObjectKey: opts.ObjectKey,
StreamID: opts.StreamID,
},
ExpiresAt: opts.ExpiresAt,
Encryption: opts.Encryption,
ZombieDeletionDeadline: opts.ZombieDeletionDeadline,
}
if err := db.db.QueryRowContext(ctx, `
INSERT INTO objects (
project_id, bucket_name, object_key, version, stream_id,
expires_at, encryption,
zombie_deletion_deadline,
encrypted_metadata, encrypted_metadata_nonce, encrypted_metadata_encrypted_key
) VALUES (
$1, $2, $3,
coalesce((
SELECT version + 1
FROM objects
WHERE project_id = $1 AND bucket_name = $2 AND object_key = $3
ORDER BY version DESC
LIMIT 1
), 1),
$4, $5, $6,
$7,
$8, $9, $10)
RETURNING status, version, created_at
`, opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.StreamID,
opts.ExpiresAt, encryptionParameters{&opts.Encryption},
opts.ZombieDeletionDeadline,
opts.EncryptedMetadata, opts.EncryptedMetadataNonce, opts.EncryptedMetadataEncryptedKey,
).Scan(&object.Status, &object.Version, &object.CreatedAt); err != nil {
return Object{}, Error.New("unable to insert object: %w", err)
}
mon.Meter("object_begin").Mark(1)
return object, nil
}
// BeginObjectExactVersion contains arguments necessary for starting an object upload.
type BeginObjectExactVersion struct {
ObjectStream
ExpiresAt *time.Time
ZombieDeletionDeadline *time.Time
EncryptedMetadata []byte // optional
EncryptedMetadataNonce []byte // optional
EncryptedMetadataEncryptedKey []byte // optional
Encryption storj.EncryptionParameters
}
// Verify verifies get object reqest fields.
func (opts *BeginObjectExactVersion) Verify() error {
if err := opts.ObjectStream.Verify(); err != nil {
return err
}
if opts.Version == NextVersion {
return ErrInvalidRequest.New("Version should not be metabase.NextVersion")
}
if opts.EncryptedMetadata == nil && (opts.EncryptedMetadataNonce != nil || opts.EncryptedMetadataEncryptedKey != nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be not set if EncryptedMetadata is not set")
} else if opts.EncryptedMetadata != nil && (opts.EncryptedMetadataNonce == nil || opts.EncryptedMetadataEncryptedKey == nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be set if EncryptedMetadata is set")
}
return nil
}
// BeginObjectExactVersion adds a pending object to the database, with specific version.
func (db *DB) BeginObjectExactVersion(ctx context.Context, opts BeginObjectExactVersion) (committed Object, err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.Verify(); err != nil {
return Object{}, err
}
if opts.ZombieDeletionDeadline == nil {
deadline := time.Now().Add(defaultZombieDeletionPeriod)
opts.ZombieDeletionDeadline = &deadline
}
object := Object{
ObjectStream: ObjectStream{
ProjectID: opts.ProjectID,
BucketName: opts.BucketName,
ObjectKey: opts.ObjectKey,
Version: opts.Version,
StreamID: opts.StreamID,
},
ExpiresAt: opts.ExpiresAt,
Encryption: opts.Encryption,
ZombieDeletionDeadline: opts.ZombieDeletionDeadline,
}
err = db.db.QueryRowContext(ctx, `
INSERT INTO objects (
project_id, bucket_name, object_key, version, stream_id,
expires_at, encryption,
zombie_deletion_deadline,
encrypted_metadata, encrypted_metadata_nonce, encrypted_metadata_encrypted_key
) VALUES (
$1, $2, $3, $4, $5,
$6, $7,
$8,
$9, $10, $11
)
RETURNING status, created_at
`, opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.Version, opts.StreamID,
opts.ExpiresAt, encryptionParameters{&opts.Encryption},
opts.ZombieDeletionDeadline,
opts.EncryptedMetadata, opts.EncryptedMetadataNonce, opts.EncryptedMetadataEncryptedKey,
).Scan(
&object.Status, &object.CreatedAt,
)
if err != nil {
if code := pgerrcode.FromError(err); code == pgxerrcode.UniqueViolation {
return Object{}, Error.Wrap(ErrObjectAlreadyExists.New(""))
}
return Object{}, Error.New("unable to insert object: %w", err)
}
mon.Meter("object_begin").Mark(1)
return object, nil
}
// BeginSegment contains options to verify, whether a new segment upload can be started.
type BeginSegment struct {
ObjectStream
Position SegmentPosition
// TODO: unused field, can remove
RootPieceID storj.PieceID
Pieces Pieces
}
// BeginSegment verifies, whether a new segment upload can be started.
func (db *DB) BeginSegment(ctx context.Context, opts BeginSegment) (err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.ObjectStream.Verify(); err != nil {
return err
}
if err := opts.Pieces.Verify(); err != nil {
return err
}
if opts.RootPieceID.IsZero() {
return ErrInvalidRequest.New("RootPieceID missing")
}
// NOTE: this isn't strictly necessary, since we can also fail this in CommitSegment.
// however, we should prevent creating segements for non-partial objects.
// Verify that object exists and is partial.
var value int
err = db.db.QueryRowContext(ctx, `
SELECT 1
FROM objects WHERE
project_id = $1 AND
bucket_name = $2 AND
object_key = $3 AND
version = $4 AND
stream_id = $5 AND
status = `+pendingStatus,
opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.Version, opts.StreamID).Scan(&value)
if err != nil {
if errors.Is(err, sql.ErrNoRows) {
return ErrPendingObjectMissing.New("")
}
return Error.New("unable to query object status: %w", err)
}
mon.Meter("segment_begin").Mark(1)
return nil
}
// CommitSegment contains all necessary information about the segment.
type CommitSegment struct {
ObjectStream
Position SegmentPosition
RootPieceID storj.PieceID
ExpiresAt *time.Time
EncryptedKeyNonce []byte
EncryptedKey []byte
PlainOffset int64 // offset in the original data stream
PlainSize int32 // size before encryption
EncryptedSize int32 // segment size after encryption
EncryptedETag []byte
Redundancy storj.RedundancyScheme
Pieces Pieces
Placement storj.PlacementConstraint
}
// CommitSegment commits segment to the database.
func (db *DB) CommitSegment(ctx context.Context, opts CommitSegment) (err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.ObjectStream.Verify(); err != nil {
return err
}
if err := opts.Pieces.Verify(); err != nil {
return err
}
switch {
case opts.RootPieceID.IsZero():
return ErrInvalidRequest.New("RootPieceID missing")
case len(opts.EncryptedKey) == 0:
return ErrInvalidRequest.New("EncryptedKey missing")
case len(opts.EncryptedKeyNonce) == 0:
return ErrInvalidRequest.New("EncryptedKeyNonce missing")
case opts.EncryptedSize <= 0:
return ErrInvalidRequest.New("EncryptedSize negative or zero")
case opts.PlainSize <= 0 && validatePlainSize:
return ErrInvalidRequest.New("PlainSize negative or zero")
case opts.PlainOffset < 0:
return ErrInvalidRequest.New("PlainOffset negative")
case opts.Redundancy.IsZero():
return ErrInvalidRequest.New("Redundancy zero")
}
if len(opts.Pieces) < int(opts.Redundancy.OptimalShares) {
return ErrInvalidRequest.New("number of pieces is less than redundancy optimal shares value")
}
aliasPieces, err := db.aliasCache.EnsurePiecesToAliases(ctx, opts.Pieces)
if err != nil {
return Error.New("unable to convert pieces to aliases: %w", err)
}
// Verify that object exists and is partial.
_, err = db.db.ExecContext(ctx, `
INSERT INTO segments (
stream_id, position, expires_at,
root_piece_id, encrypted_key_nonce, encrypted_key,
encrypted_size, plain_offset, plain_size, encrypted_etag,
redundancy,
remote_alias_pieces,
placement
) VALUES (
(SELECT stream_id
FROM objects WHERE
project_id = $12 AND
bucket_name = $13 AND
object_key = $14 AND
version = $15 AND
stream_id = $16 AND
status = `+pendingStatus+
` ), $1, $2,
$3, $4, $5,
$6, $7, $8, $9,
$10,
$11,
$17
)
ON CONFLICT(stream_id, position)
DO UPDATE SET
expires_at = $2,
root_piece_id = $3, encrypted_key_nonce = $4, encrypted_key = $5,
encrypted_size = $6, plain_offset = $7, plain_size = $8, encrypted_etag = $9,
redundancy = $10,
remote_alias_pieces = $11,
placement = $17
`, opts.Position, opts.ExpiresAt,
opts.RootPieceID, opts.EncryptedKeyNonce, opts.EncryptedKey,
opts.EncryptedSize, opts.PlainOffset, opts.PlainSize, opts.EncryptedETag,
redundancyScheme{&opts.Redundancy},
aliasPieces,
opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.Version, opts.StreamID,
opts.Placement,
)
if err != nil {
if code := pgerrcode.FromError(err); code == pgxerrcode.NotNullViolation {
return ErrPendingObjectMissing.New("")
}
return Error.New("unable to insert segment: %w", err)
}
mon.Meter("segment_commit").Mark(1)
mon.IntVal("segment_commit_encrypted_size").Observe(int64(opts.EncryptedSize))
return nil
}
// CommitInlineSegment contains all necessary information about the segment.
type CommitInlineSegment struct {
ObjectStream
Position SegmentPosition
ExpiresAt *time.Time
EncryptedKeyNonce []byte
EncryptedKey []byte
PlainOffset int64 // offset in the original data stream
PlainSize int32 // size before encryption
EncryptedETag []byte
InlineData []byte
}
// CommitInlineSegment commits inline segment to the database.
func (db *DB) CommitInlineSegment(ctx context.Context, opts CommitInlineSegment) (err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.ObjectStream.Verify(); err != nil {
return err
}
// TODO: do we have a lower limit for inline data?
// TODO should we move check for max inline segment from metainfo here
switch {
case len(opts.EncryptedKey) == 0:
return ErrInvalidRequest.New("EncryptedKey missing")
case len(opts.EncryptedKeyNonce) == 0:
return ErrInvalidRequest.New("EncryptedKeyNonce missing")
case opts.PlainSize <= 0 && validatePlainSize:
return ErrInvalidRequest.New("PlainSize negative or zero")
case opts.PlainOffset < 0:
return ErrInvalidRequest.New("PlainOffset negative")
}
// Verify that object exists and is partial.
_, err = db.db.ExecContext(ctx, `
INSERT INTO segments (
stream_id, position, expires_at,
root_piece_id, encrypted_key_nonce, encrypted_key,
encrypted_size, plain_offset, plain_size, encrypted_etag,
inline_data
) VALUES (
(SELECT stream_id
FROM objects WHERE
project_id = $11 AND
bucket_name = $12 AND
object_key = $13 AND
version = $14 AND
stream_id = $15 AND
status = `+pendingStatus+
` ), $1, $2,
$3, $4, $5,
$6, $7, $8, $9,
$10
)
ON CONFLICT(stream_id, position)
DO UPDATE SET
expires_at = $2,
root_piece_id = $3, encrypted_key_nonce = $4, encrypted_key = $5,
encrypted_size = $6, plain_offset = $7, plain_size = $8, encrypted_etag = $9,
inline_data = $10
`, opts.Position, opts.ExpiresAt,
storj.PieceID{}, opts.EncryptedKeyNonce, opts.EncryptedKey,
len(opts.InlineData), opts.PlainOffset, opts.PlainSize, opts.EncryptedETag,
opts.InlineData,
opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.Version, opts.StreamID,
)
if err != nil {
if code := pgerrcode.FromError(err); code == pgxerrcode.NotNullViolation {
return ErrPendingObjectMissing.New("")
}
return Error.New("unable to insert segment: %w", err)
}
mon.Meter("segment_commit").Mark(1)
mon.IntVal("segment_commit_encrypted_size").Observe(int64(len(opts.InlineData)))
return nil
}
// CommitObject contains arguments necessary for committing an object.
type CommitObject struct {
ObjectStream
Encryption storj.EncryptionParameters
// this flag controls if we want to set metadata fields with CommitObject
// it's possible to set metadata with BeginObject request so we need to
// be explicit if we would like to set it with CommitObject which will
// override any existing metadata.
OverrideEncryptedMetadata bool
EncryptedMetadata []byte // optional
EncryptedMetadataNonce []byte // optional
EncryptedMetadataEncryptedKey []byte // optional
DisallowDelete bool
}
// Verify verifies reqest fields.
func (c *CommitObject) Verify() error {
if err := c.ObjectStream.Verify(); err != nil {
return err
}
if c.Encryption.CipherSuite != storj.EncUnspecified && c.Encryption.BlockSize <= 0 {
return ErrInvalidRequest.New("Encryption.BlockSize is negative or zero")
}
if c.OverrideEncryptedMetadata {
if c.EncryptedMetadata == nil && (c.EncryptedMetadataNonce != nil || c.EncryptedMetadataEncryptedKey != nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be not set if EncryptedMetadata is not set")
} else if c.EncryptedMetadata != nil && (c.EncryptedMetadataNonce == nil || c.EncryptedMetadataEncryptedKey == nil) {
return ErrInvalidRequest.New("EncryptedMetadataNonce and EncryptedMetadataEncryptedKey must be set if EncryptedMetadata is set")
}
}
return nil
}
// CommitObject adds a pending object to the database.
func (db *DB) CommitObject(ctx context.Context, opts CommitObject) (object Object, err error) {
defer mon.Task()(&ctx)(&err)
if err := opts.Verify(); err != nil {
return Object{}, err
}
err = txutil.WithTx(ctx, db.db, nil, func(ctx context.Context, tx tagsql.Tx) error {
segments, err := fetchSegmentsForCommit(ctx, tx, opts.StreamID)
if err != nil {
return Error.New("failed to fetch segments: %w", err)
}
if err = db.validateParts(segments); err != nil {
return err
}
finalSegments := convertToFinalSegments(segments)
err = updateSegmentOffsets(ctx, tx, opts.StreamID, finalSegments)
if err != nil {
return Error.New("failed to update segments: %w", err)
}
// TODO: would we even need this when we make main index plain_offset?
fixedSegmentSize := int32(0)
if len(finalSegments) > 0 {
fixedSegmentSize = finalSegments[0].PlainSize
for i, seg := range finalSegments {
if seg.Position.Part != 0 || seg.Position.Index != uint32(i) {
fixedSegmentSize = -1
break
}
if i < len(finalSegments)-1 && seg.PlainSize != fixedSegmentSize {
fixedSegmentSize = -1
break
}
}
}
var totalPlainSize, totalEncryptedSize int64
for _, seg := range finalSegments {
totalPlainSize += int64(seg.PlainSize)
totalEncryptedSize += int64(seg.EncryptedSize)
}
args := []interface{}{
opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey, opts.Version, opts.StreamID,
len(segments),
totalPlainSize,
totalEncryptedSize,
fixedSegmentSize,
encryptionParameters{&opts.Encryption},
}
metadataColumns := ""
if opts.OverrideEncryptedMetadata {
args = append(args,
opts.EncryptedMetadataNonce,
opts.EncryptedMetadata,
opts.EncryptedMetadataEncryptedKey,
)
metadataColumns = `,
encrypted_metadata_nonce = $11,
encrypted_metadata = $12,
encrypted_metadata_encrypted_key = $13
`
}
versionsToDelete := []Version{}
if db.config.MultipleVersions {
if err := withRows(tx.QueryContext(ctx, `
SELECT version
FROM objects
WHERE
project_id = $1 AND
bucket_name = $2 AND
object_key = $3 AND
status = `+committedStatus,
opts.ProjectID, []byte(opts.BucketName), opts.ObjectKey))(func(rows tagsql.Rows) error {
for rows.Next() {
var version Version
if err := rows.Scan(&version); err != nil {
return Error.New("failed to scan previous object: %w", err)
}
versionsToDelete = append(versionsToDelete, version)
}
return nil
}); err != nil {
return Error.New("failed to find previous objects: %w", err)
}
if len(versionsToDelete) > 1 {
db.log.Warn("object with multiple committed versions were found!",
zap.Stringer("Project ID", opts.ProjectID), zap.String("Bucket Name", opts.BucketName),
zap.String("Object Key", string(opts.ObjectKey)))
mon.Meter("multiple_committed_versions").Mark(1)
}
if len(versionsToDelete) != 0 && opts.DisallowDelete {
return ErrPermissionDenied.New("no permissions to delete existing object")
}
}
err = tx.QueryRowContext(ctx, `
UPDATE objects SET
status =`+committedStatus+`,
segment_count = $6,
total_plain_size = $7,
total_encrypted_size = $8,
fixed_segment_size = $9,
zombie_deletion_deadline = NULL,
-- TODO should we allow to override existing encryption parameters or return error if don't match with opts?
encryption = CASE
WHEN objects.encryption = 0 AND $10 <> 0 THEN $10
WHEN objects.encryption = 0 AND $10 = 0 THEN NULL
ELSE objects.encryption
END
`+metadataColumns+`
WHERE
project_id = $1 AND
bucket_name = $2 AND
object_key = $3 AND
version = $4 AND
stream_id = $5 AND
status = `+pendingStatus+`
RETURNING
created_at, expires_at,
encrypted_metadata, encrypted_metadata_encrypted_key, encrypted_metadata_nonce,
encryption
`, args...).Scan(
&object.CreatedAt, &object.ExpiresAt,
&object.EncryptedMetadata, &object.EncryptedMetadataEncryptedKey, &object.EncryptedMetadataNonce,
encryptionParameters{&object.Encryption},
)
if err != nil {
if errors.Is(err, sql.ErrNoRows) {
return storj.ErrObjectNotFound.Wrap(Error.New("object with specified version and pending status is missing"))
} else if code := pgerrcode.FromError(err); code == pgxerrcode.NotNullViolation {
// TODO maybe we should check message if 'encryption' label is there
return ErrInvalidRequest.New("Encryption is missing")
}
return Error.New("failed to update object: %w", err)
}
for _, version := range versionsToDelete {
// TODO delete pieces from stoage nodes
_, err := db.deleteObjectExactVersion(ctx, DeleteObjectExactVersion{
ObjectLocation: ObjectLocation{
ProjectID: opts.ProjectID,
BucketName: opts.BucketName,
ObjectKey: opts.ObjectKey,
},
Version: version,
}, tx)
if err != nil {
return Error.New("failed to delete existing object: %w", err)
}
}
object.StreamID = opts.StreamID
object.ProjectID = opts.ProjectID
object.BucketName = opts.BucketName
object.ObjectKey = opts.ObjectKey
object.Version = opts.Version
object.Status = Committed
object.SegmentCount = int32(len(segments))
object.TotalPlainSize = totalPlainSize
object.TotalEncryptedSize = totalEncryptedSize
object.FixedSegmentSize = fixedSegmentSize
return nil
})
if err != nil {
return Object{}, err
}
mon.Meter("object_commit").Mark(1)
mon.IntVal("object_commit_segments").Observe(int64(object.SegmentCount))
mon.IntVal("object_commit_encrypted_size").Observe(object.TotalEncryptedSize)
return object, nil
}
func (db *DB) validateParts(segments []segmentInfoForCommit) error {
partSize := make(map[uint32]memory.Size)
var lastPart uint32
for _, segment := range segments {
partSize[segment.Position.Part] += memory.Size(segment.PlainSize)
if lastPart < segment.Position.Part {
lastPart = segment.Position.Part
}
}
if len(partSize) > db.config.MaxNumberOfParts {
return Error.New("exceeded maximum number of parts: %d", db.config.MaxNumberOfParts)
}
for part, size := range partSize {
// Last part has no minimum size.
if part == lastPart {
continue
}
if size < db.config.MinPartSize {
return Error.New("size of part number %d is below minimum threshold, got: %s, min: %s", part, size, db.config.MinPartSize)
}
}
return nil
}