2014-11-18 00:29:42 +00:00
|
|
|
package storage
|
|
|
|
|
|
|
|
import (
|
|
|
|
"fmt"
|
|
|
|
"io"
|
|
|
|
"time"
|
2014-11-19 22:39:32 +00:00
|
|
|
|
2014-12-24 00:01:38 +00:00
|
|
|
"github.com/docker/distribution/digest"
|
2015-01-02 21:21:29 +00:00
|
|
|
"github.com/docker/distribution/manifest"
|
2014-11-18 00:29:42 +00:00
|
|
|
)
|
|
|
|
|
|
|
|
// Layer provides a readable and seekable layer object. Typically,
|
|
|
|
// implementations are *not* goroutine safe.
|
|
|
|
type Layer interface {
|
|
|
|
// http.ServeContent requires an efficient implementation of
|
|
|
|
// ReadSeeker.Seek(0, os.SEEK_END).
|
|
|
|
io.ReadSeeker
|
|
|
|
io.Closer
|
|
|
|
|
|
|
|
// Name returns the repository under which this layer is linked.
|
|
|
|
Name() string // TODO(stevvooe): struggling with nomenclature: should this be "repo" or "name"?
|
|
|
|
|
2014-11-19 22:39:32 +00:00
|
|
|
// Digest returns the unique digest of the blob, which is the tarsum for
|
|
|
|
// layers.
|
|
|
|
Digest() digest.Digest
|
2014-11-18 00:29:42 +00:00
|
|
|
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
// CreatedAt returns the time this layer was created.
|
2014-11-18 00:29:42 +00:00
|
|
|
CreatedAt() time.Time
|
|
|
|
}
|
|
|
|
|
|
|
|
// LayerUpload provides a handle for working with in-progress uploads.
|
|
|
|
// Instances can be obtained from the LayerService.Upload and
|
|
|
|
// LayerService.Resume.
|
|
|
|
type LayerUpload interface {
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
io.WriteSeeker
|
|
|
|
io.Closer
|
2014-11-18 00:29:42 +00:00
|
|
|
|
|
|
|
// Name of the repository under which the layer will be linked.
|
|
|
|
Name() string
|
|
|
|
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
// UUID returns the identifier for this upload.
|
|
|
|
UUID() string
|
2014-11-18 00:29:42 +00:00
|
|
|
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
// StartedAt returns the time this layer upload was started.
|
|
|
|
StartedAt() time.Time
|
2014-12-13 01:43:30 +00:00
|
|
|
|
2014-11-18 00:29:42 +00:00
|
|
|
// Finish marks the upload as completed, returning a valid handle to the
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
// uploaded layer. The digest is validated against the contents of the
|
|
|
|
// uploaded layer.
|
|
|
|
Finish(digest digest.Digest) (Layer, error)
|
2014-11-18 00:29:42 +00:00
|
|
|
|
|
|
|
// Cancel the layer upload process.
|
|
|
|
Cancel() error
|
|
|
|
}
|
|
|
|
|
|
|
|
var (
|
|
|
|
// ErrLayerExists returned when layer already exists
|
|
|
|
ErrLayerExists = fmt.Errorf("layer exists")
|
|
|
|
|
|
|
|
// ErrLayerTarSumVersionUnsupported when tarsum is unsupported version.
|
|
|
|
ErrLayerTarSumVersionUnsupported = fmt.Errorf("unsupported tarsum version")
|
|
|
|
|
|
|
|
// ErrLayerUploadUnknown returned when upload is not found.
|
|
|
|
ErrLayerUploadUnknown = fmt.Errorf("layer upload unknown")
|
|
|
|
|
2014-11-21 01:49:35 +00:00
|
|
|
// ErrLayerClosed returned when an operation is attempted on a closed
|
|
|
|
// Layer or LayerUpload.
|
|
|
|
ErrLayerClosed = fmt.Errorf("layer closed")
|
2014-11-18 00:29:42 +00:00
|
|
|
)
|
2014-11-26 20:52:52 +00:00
|
|
|
|
|
|
|
// ErrUnknownLayer returned when layer cannot be found.
|
|
|
|
type ErrUnknownLayer struct {
|
2015-01-02 21:21:29 +00:00
|
|
|
FSLayer manifest.FSLayer
|
2014-11-26 20:52:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func (err ErrUnknownLayer) Error() string {
|
|
|
|
return fmt.Sprintf("unknown layer %v", err.FSLayer.BlobSum)
|
|
|
|
}
|
|
|
|
|
|
|
|
// ErrLayerInvalidDigest returned when tarsum check fails.
|
|
|
|
type ErrLayerInvalidDigest struct {
|
Spool layer uploads to remote storage
To smooth initial implementation, uploads were spooled to local file storage,
validated, then pushed to remote storage. That approach was flawed in that it
present easy clustering of registry services that share a remote storage
backend. The original plan was to implement resumable hashes then implement
remote upload storage. After some thought, it was found to be better to get
remote spooling working, then optimize with resumable hashes.
Moving to this approach has tradeoffs: after storing the complete upload
remotely, the node must fetch the content and validate it before moving it to
the final location. This can double bandwidth usage to the remote backend.
Modifying the verification and upload code to store intermediate hashes should
be trivial once the layer digest format has settled.
The largest changes for users of the storage package (mostly the registry app)
are the LayerService interface and the LayerUpload interface. The LayerService
now takes qualified repository names to start and resume uploads. In corallry,
the concept of LayerUploadState has been complete removed, exposing all aspects
of that state as part of the LayerUpload object. The LayerUpload object has
been modified to work as an io.WriteSeeker and includes a StartedAt time, to
allow for upload timeout policies. Finish now only requires a digest, eliding
the requirement for a size parameter.
Resource cleanup has taken a turn for the better. Resources are cleaned up
after successful uploads and during a cancel call. Admittedly, this is probably
not completely where we want to be. It's recommend that we bolster this with a
periodic driver utility script that scans for partial uploads and deletes the
underlying data. As a small benefit, we can leave these around to better
understand how and why these uploads are failing, at the cost of some extra
disk space.
Many other changes follow from the changes above. The webapp needs to be
updated to meet the new interface requirements.
Signed-off-by: Stephen J Day <stephen.day@docker.com>
2015-01-08 22:24:02 +00:00
|
|
|
Digest digest.Digest
|
2015-02-02 21:01:49 +00:00
|
|
|
Reason error
|
2014-11-26 20:52:52 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
func (err ErrLayerInvalidDigest) Error() string {
|
2015-02-02 21:01:49 +00:00
|
|
|
return fmt.Sprintf("invalid digest for referenced layer: %v, %v",
|
|
|
|
err.Digest, err.Reason)
|
2015-01-30 04:45:19 +00:00
|
|
|
}
|