-
Notifications
You must be signed in to change notification settings - Fork 6
/
builder.go
542 lines (477 loc) · 14.7 KB
/
builder.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
/*
Copyright The starlight Authors.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
file created by maverick in 2021
*/
package proxy
import (
"bytes"
"compress/gzip"
"database/sql"
"encoding/json"
"fmt"
"github.com/containerd/containerd/log"
"github.com/google/go-containerregistry/pkg/name"
"github.com/mc256/starlight/util"
"github.com/mc256/starlight/util/common"
"github.com/mc256/starlight/util/send"
"github.com/opencontainers/go-digest"
"github.com/pkg/errors"
"golang.org/x/sync/errgroup"
"io"
"math"
"net/http"
"sort"
)
////////////////////////////////////////////////
type Builder struct {
send.DeltaBundle
// non-exported fields
server *Server
manifest, config []byte
manifestDigest string
unavailableLayers, availableLayers []*send.ImageLayer
}
func (b *Builder) String() string {
if b.Source != nil {
return fmt.Sprintf("Builder (%s)->(%s)", b.Source.Ref.String(), b.Destination.Ref.String())
}
return fmt.Sprintf("Builder ()->(%s)", b.Destination.Ref.String())
}
func (b *Builder) WriteHeader(w http.ResponseWriter, req *http.Request) error {
buf := bytes.NewBuffer(make([]byte, 0))
cw := util.NewCountWriter(buf)
// manifest
gwManifest, err := gzip.NewWriterLevel(cw, gzip.BestCompression)
if err != nil {
return err
}
_, err = gwManifest.Write(b.manifest)
if err != nil {
return err
}
err = gwManifest.Close()
if err != nil {
return err
}
manifestSize := cw.GetWrittenSize()
// config
gwConfig, err := gzip.NewWriterLevel(cw, gzip.BestCompression)
if err != nil {
return err
}
_, err = gwConfig.Write(b.config)
if err != nil {
return err
}
err = gwConfig.Close()
if err != nil {
return err
}
configSize := cw.GetWrittenSize() - manifestSize
// header
h, err := json.Marshal(b)
if err != nil {
return err
}
gwHeader, err := gzip.NewWriterLevel(cw, gzip.BestCompression)
if err != nil {
return err
}
_, err = gwHeader.Write(h)
if err != nil {
return err
}
err = gwHeader.Close()
if err != nil {
return err
}
headerSize := cw.GetWrittenSize() - manifestSize - configSize
slDigest := digest.FromBytes(h)
log.G(b.server.ctx).
WithField("manifest", manifestSize).
WithField("config", configSize).
WithField("header", headerSize).
WithField("body", b.BodyLength).
WithField("_digest", slDigest.String()).
Info("generated response header")
// output header
httpLength := cw.GetWrittenSize() + b.BodyLength
header := w.Header()
header.Set("Content-Type", "application/octet-stream")
header.Set("Content-Length", fmt.Sprintf("%d", httpLength))
header.Set("Starlight-Header-Size", fmt.Sprintf("%d", headerSize))
header.Set("Manifest-Size", fmt.Sprintf("%d", manifestSize))
header.Set("Config-Size", fmt.Sprintf("%d", configSize))
header.Set("Digest", b.manifestDigest)
header.Set("Starlight-Digest", slDigest.String())
header.Set("Starlight-Version", util.Version)
header.Set("Content-Disposition", `attachment; filename="starlight.tgz"`)
w.WriteHeader(http.StatusOK)
_, err = w.Write(buf.Bytes())
return err
}
func (b *Builder) WriteBody(w http.ResponseWriter, req *http.Request) error {
layers := b.Destination.Layers
// output body
for _, c := range b.Contents {
// fmt.Println(c.files[0].Name, len(c.Chunks), c.Stack, len(c.files), c.Offset, c.Size, layers[c.Stack].blob)
if layer := layers[c.Stack]; layer.Blob != nil {
sr := io.NewSectionReader(layer.Blob.Buffer, 0, layer.UncompressedSize)
for _, chunk := range c.Chunks {
ssr := io.NewSectionReader(sr, chunk.Offset, chunk.CompressedSize)
_, err := io.CopyN(w, ssr, chunk.CompressedSize)
if err != nil {
return errors.Wrapf(err, "failed to copy chunk at [%d] for file [%s]", chunk.Offset, c.Files[0].Name)
}
}
} else {
return fmt.Errorf("layer %d not found for file %s (ref:%d)", c.Stack, c.Files[0].Name, len(c.Files))
}
}
return nil
}
func (b *Builder) fetchLayers(cache *send.ImageLayer) error {
var (
c *common.LayerCache
has bool
)
b.server.cacheMutex.Lock()
if c, has = b.server.cache[cache.Hash]; has {
b.server.cacheMutex.Unlock()
sub := make(chan error)
c.Subscribe(&sub)
err := <-sub
if err != nil {
log.G(b.server.ctx).
WithField("layer", cache.String()).
Error(errors.Wrapf(err, "failed to load layer"))
return err
}
log.G(b.server.ctx).
WithField("layer", cache.String()).
WithField("shared", true).
Info("fetched layer")
} else {
c = common.NewLayerCache(cache)
b.server.cache[cache.Hash] = c
b.server.cacheMutex.Unlock()
if err := c.Load(b.server.ctx); err != nil {
log.G(b.server.ctx).
WithField("layer", cache.String()).
Error(errors.Wrapf(err, "failed to load layer"))
return err
}
log.G(b.server.ctx).
WithField("layer", cache.String()).
WithField("shared", false).
Info("fetched layer")
}
cache.Blob = c
return nil
}
// getImage returns the image with the given reference and the platform.
// if you need to find out the available image, you should better use getImageByDigest which returns
// the exact image that is available as tags might be changed but the digest will not change.
func (b *Builder) getImage(ref, platform string) (img *send.Image, err error) {
img = &send.Image{}
img.Ref, err = name.ParseReference(ref,
name.WithDefaultRegistry(b.server.config.DefaultRegistry),
name.WithDefaultTag("latest-starlight"),
)
if err != nil {
return nil, err
}
refName, refTag := ParseImageReference(img.Ref, b.server.config.DefaultRegistry)
img.Serial, err = b.server.db.GetImage(refName, refTag, platform)
if err != nil {
return nil, err
}
// Load necessary layers
img.Layers, err = b.server.db.GetLayers(img.Serial)
if err != nil {
return nil, err
}
for _, layer := range img.Layers {
d := fmt.Sprintf("%s@%s", img.Ref.Name(), layer.Hash)
var dd name.Digest
dd, err = name.NewDigest(d)
layer.SetDigest(dd)
}
return img, nil
}
// getImageByDigest returns a more precise image reference by using digest (not tag)
// this guarantees that the image is the exact image that is available on the client side.
func (b Builder) getImageByDigest(refWithDigest string) (img *send.Image, err error) {
img = &send.Image{}
img.Ref, err = name.ParseReference(refWithDigest,
name.WithDefaultRegistry(b.server.config.DefaultRegistry),
name.WithDefaultTag("latest-starlight"),
)
if err != nil {
return nil, err
}
refName, refTag := ParseImageReference(img.Ref, b.server.config.DefaultRegistry)
img.Serial, err = b.server.db.GetImageByDigest(refName, refTag)
if err != nil {
return nil, err
}
// Load necessary layers
img.Layers, err = b.server.db.GetLayers(img.Serial)
if err != nil {
return nil, err
}
for _, layer := range img.Layers {
d := fmt.Sprintf("%s@%s", img.Ref.Name(), layer.Hash)
var dd name.Digest
dd, err = name.NewDigest(d)
layer.SetDigest(dd)
}
return img, nil
}
// getManifestAndConfig returns the manifest and config of the image by its serial number.
// This is important for creating an valid container image on the client side.
func (b Builder) getManifestAndConfig(serial int64) (config, manifest []byte, digest string, err error) {
return b.server.db.GetManifestAndConfig(serial)
}
func (b *Builder) getUnavailableLayers() (layers []*send.ImageLayer, err error) {
dedup := make(map[string]bool, len(b.Destination.Layers))
layers = make([]*send.ImageLayer, 0, len(b.Destination.Layers))
if b.Source != nil {
for _, v := range b.Source.Layers {
dedup[v.Hash] = true
}
}
for _, v := range b.Destination.Layers {
if _, has := dedup[v.Hash]; !has {
layers = append(layers, v)
} else {
dedup[v.Hash] = true
}
}
return layers, nil
}
func (b *Builder) computeDelta() error {
/*
1. compute the set of existing files from available layers
2. compute the set of requested files from non-existing layers
3. identify the best reference to the file content
4. sort file contents
*/
unavailable, available := make([]*send.ImageLayer, 0), make([]*send.ImageLayer, 0)
availableIds := make(map[int64]send.LayerSource)
if b.Source != nil {
for _, layer := range b.Source.Layers {
if layer.Available {
available = append(available, layer)
availableIds[layer.Serial] = send.FromSource
} else {
unavailable = append(unavailable, layer)
}
}
}
for _, layer := range b.Destination.Layers {
if layer.Available {
available = append(available, layer)
availableIds[layer.Serial] = send.FromDestination
} else {
unavailable = append(unavailable, layer)
}
}
// 1. compute the set of existing files from available layers
existingFiles, err := b.server.db.GetUniqueFiles(available)
if err != nil {
return errors.Wrapf(err, "failed to get existing files")
}
deduplicatedExistingFiles := make(map[string]*send.File)
for _, f := range existingFiles {
// find the lowest layer, if available find it in the destination image, so we can create references within the
// requested image and reduce future cleanup steps
if fe, has := deduplicatedExistingFiles[f.Digest]; has && ((f.FsId < fe.FsId) || (availableIds[f.FsId] == send.FromDestination)) {
continue
}
deduplicatedExistingFiles[f.Digest] = f
}
log.G(b.server.ctx).
WithField("unique", len(deduplicatedExistingFiles)).
WithField("total", len(existingFiles)).
WithField("builder", b).
WithField("_step", 1).
Info("find existing file contents")
// 2. compute the set of requested files from non-existing layers
b.RequestedFiles, err = b.server.db.GetFilesWithRanks(b.Destination.Serial)
if err != nil {
return errors.Wrapf(err, "failed to get requested files")
}
deduplicatedRequestedContents := make(map[string]*send.Content)
for _, f := range b.RequestedFiles {
if f.Digest == "" {
continue
}
if f.Size == 0 {
continue
}
if _, has := availableIds[f.FsId]; has {
continue
}
if fe, has := deduplicatedRequestedContents[f.Digest]; has {
fe.Files = append(fe.Files, f)
continue
}
deduplicatedRequestedContents[f.Digest] = &send.Content{
Files: []*send.RankedFile{f},
Rank: 0,
}
}
log.G(b.server.ctx).
WithField("unique", len(deduplicatedRequestedContents)).
WithField("total", len(b.RequestedFiles)).
WithField("builder", b).
WithField("_step", 2).
Info("find requested file contents")
// 3. identify the best reference to the file content
b.Contents = make([]*send.Content, 0)
for d, reqContent := range deduplicatedRequestedContents {
if existingContent, has := deduplicatedExistingFiles[d]; has {
// found requested file in existing files
for _, r := range reqContent.Files {
r.ReferenceFsId = existingContent.FsId
}
} else {
// not found, add to the list of contents to be sent to the client
b.Contents = append(b.Contents, reqContent)
reqContent.Rank = math.MaxFloat64
reqContent.Stack = reqContent.Files[0].Stack
for _, f := range reqContent.Files {
if f.Rank < reqContent.Rank {
// highest rank wins
reqContent.Rank = f.Rank
}
if f.Stack < reqContent.Stack {
// lowest stack wins
reqContent.Stack = f.Stack
}
}
}
}
sort.Sort(send.ByRank(b.Contents))
b.BodyLength = int64(0)
for idx, c := range b.Contents {
// update the reference to the content
for _, f := range c.Files {
f.ReferenceStack = c.Stack
f.PayloadOrder = idx
}
c.Offset = b.BodyLength
c.Digest = c.Files[0].Digest
if len(c.Files[0].ParsingChunks) > 0 {
c.Chunks = make([]*send.FileChunk, 0, len(c.Files[0].ParsingChunks))
for _, chunk := range c.Files[0].ParsingChunks {
c.Size += chunk.CompressedSize
c.Chunks = append(c.Chunks, &send.FileChunk{
Offset: chunk.Offset,
ChunkOffset: chunk.ChunkOffset,
ChunkSize: chunk.ChunkSize,
CompressedSize: chunk.CompressedSize,
})
}
} else {
c.Size = c.Files[0].CompressedSize
c.Chunks = []*send.FileChunk{{
Offset: c.Files[0].Offset,
ChunkOffset: c.Files[0].ChunkOffset,
ChunkSize: c.Files[0].ChunkSize,
CompressedSize: c.Files[0].CompressedSize,
}}
}
// calculate the total size of the compressed contents
b.BodyLength += c.Size
}
log.G(b.server.ctx).
WithField("content", len(b.Contents)).
WithField("builder", b).
WithField("compressedSize", b.BodyLength).
WithField("_step", 3).
Info("find the best file content references")
return nil
}
func (b *Builder) Load() error {
// Load compressed layers from registry
var errGrp errgroup.Group
for _, layer := range b.unavailableLayers {
layer := layer
errGrp.Go(func() error {
return b.fetchLayers(layer)
})
}
// Load manifest and config from proxy's database
errGrp.Go(func() error {
if c, m, d, err := b.getManifestAndConfig(b.Destination.Serial); err != nil {
return err
} else {
b.config = c
b.manifest = m
b.manifestDigest = d
return nil
}
})
// Computer the difference between the requested and existing files
errGrp.Go(b.computeDelta)
// Wait for all jobs to end
if err := errGrp.Wait(); err != nil {
return errors.Wrapf(err, "failed to load all compressed layer")
}
return nil
}
func NewBuilder(server *Server, src, dst, plt string) (b *Builder, err error) {
b = &Builder{
server: server,
}
// Available Image:
// It must be specified by the unique hash of the image, we don't want to refercing something that is
// unknown (for security reasons). Therefore, platform is not required.
if src != "" {
if b.Source, err = b.getImageByDigest(src); err != nil {
if err == sql.ErrNoRows {
return nil, fmt.Errorf("source image %s not found", src)
} else {
return nil, errors.Wrapf(err, "failed to get source image")
}
}
}
// Requested Image:
// requires image name tag as well as the platform to be specified
// you could not download an image index for that because only one platform is needed to run the container
if b.Destination, err = b.getImage(dst, plt); err != nil {
if err == sql.ErrNoRows {
return nil, fmt.Errorf("requested image %s not found", dst)
} else {
return nil, errors.Wrapf(err, "failed to obtain requested image")
}
}
if b.Source != nil {
b.availableLayers = b.Source.Layers
b.unavailableLayers, err = b.getUnavailableLayers()
if err != nil {
return nil, err
}
} else {
b.availableLayers = []*send.ImageLayer{}
b.unavailableLayers = b.Destination.Layers
}
for _, a := range b.availableLayers {
a.Available = true
}
for _, u := range b.unavailableLayers {
u.Available = false
}
return b, nil
}