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
|
package filesys
import (
"bytes"
"context"
"fmt"
"time"
"github.com/chrislusf/seaweedfs/weed/glog"
"github.com/chrislusf/seaweedfs/weed/operation"
"github.com/chrislusf/seaweedfs/weed/pb/filer_pb"
"sync"
)
type ContinuousDirtyPages struct {
hasData bool
Offset int64
Size int64
Data []byte
f *File
lock sync.Mutex
}
func newDirtyPages(file *File) *ContinuousDirtyPages {
return &ContinuousDirtyPages{
Data: make([]byte, file.wfs.option.ChunkSizeLimit),
f: file,
}
}
func (pages *ContinuousDirtyPages) AddPage(ctx context.Context, offset int64, data []byte) (chunks []*filer_pb.FileChunk, err error) {
pages.lock.Lock()
defer pages.lock.Unlock()
var chunk *filer_pb.FileChunk
if len(data) > len(pages.Data) {
// this is more than what buffer can hold.
return pages.flushAndSave(ctx, offset, data)
}
if offset < pages.Offset || offset >= pages.Offset+int64(len(pages.Data)) ||
pages.Offset+int64(len(pages.Data)) < offset+int64(len(data)) {
// if the data is out of range,
// or buffer is full if adding new data,
// flush current buffer and add new data
// println("offset", offset, "size", len(data), "existing offset", pages.Offset, "size", pages.Size)
if chunk, err = pages.saveExistingPagesToStorage(ctx); err == nil {
if chunk != nil {
glog.V(4).Infof("%s/%s add save [%d,%d)", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size))
chunks = append(chunks, chunk)
}
} else {
glog.V(0).Infof("%s/%s add save [%d,%d): %v", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size), err)
return
}
pages.Offset = offset
copy(pages.Data, data)
pages.Size = int64(len(data))
return
}
if offset != pages.Offset+pages.Size {
// when this happens, debug shows the data overlapping with existing data is empty
// the data is not just append
if offset == pages.Offset && int(pages.Size) < len(data) {
// glog.V(2).Infof("pages[%d,%d) pages.Data len=%v, data len=%d, pages.Size=%d", pages.Offset, pages.Offset+pages.Size, len(pages.Data), len(data), pages.Size)
copy(pages.Data[pages.Size:], data[pages.Size:])
} else {
if pages.Size != 0 {
glog.V(0).Infof("possible error: pages [%d, %d) write [%d, %d)", pages.Offset, pages.Offset+pages.Size, offset, offset+int64(len(data)))
}
return pages.flushAndSave(ctx, offset, data)
}
} else {
copy(pages.Data[offset-pages.Offset:], data)
}
pages.Size = max(pages.Size, offset+int64(len(data))-pages.Offset)
return
}
func (pages *ContinuousDirtyPages) flushAndSave(ctx context.Context, offset int64, data []byte) (chunks []*filer_pb.FileChunk, err error) {
var chunk *filer_pb.FileChunk
// flush existing
if chunk, err = pages.saveExistingPagesToStorage(ctx); err == nil {
if chunk != nil {
glog.V(4).Infof("%s/%s flush existing [%d,%d)", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size))
chunks = append(chunks, chunk)
}
} else {
glog.V(0).Infof("%s/%s failed to flush1 [%d,%d): %v", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size), err)
return
}
pages.Size = 0
pages.Offset = 0
// flush the new page
if chunk, err = pages.saveToStorage(ctx, data, offset); err == nil {
if chunk != nil {
glog.V(4).Infof("%s/%s flush big request [%d,%d)", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size))
chunks = append(chunks, chunk)
}
} else {
glog.V(0).Infof("%s/%s failed to flush2 [%d,%d): %v", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size), err)
return
}
return
}
func (pages *ContinuousDirtyPages) FlushToStorage(ctx context.Context) (chunk *filer_pb.FileChunk, err error) {
pages.lock.Lock()
defer pages.lock.Unlock()
if pages.Size == 0 {
return nil, nil
}
if chunk, err = pages.saveExistingPagesToStorage(ctx); err == nil {
pages.Size = 0
pages.Offset = 0
if chunk != nil {
glog.V(4).Infof("%s/%s flush [%d,%d)", pages.f.dir.Path, pages.f.Name, chunk.Offset, chunk.Offset+int64(chunk.Size))
}
}
return
}
func (pages *ContinuousDirtyPages) saveExistingPagesToStorage(ctx context.Context) (*filer_pb.FileChunk, error) {
if pages.Size == 0 {
return nil, nil
}
return pages.saveToStorage(ctx, pages.Data[:pages.Size], pages.Offset)
}
func (pages *ContinuousDirtyPages) saveToStorage(ctx context.Context, buf []byte, offset int64) (*filer_pb.FileChunk, error) {
var fileId, host string
if err := pages.f.wfs.withFilerClient(func(client filer_pb.SeaweedFilerClient) error {
request := &filer_pb.AssignVolumeRequest{
Count: 1,
Replication: pages.f.wfs.option.Replication,
Collection: pages.f.wfs.option.Collection,
TtlSec: pages.f.wfs.option.TtlSec,
DataCenter: pages.f.wfs.option.DataCenter,
}
resp, err := client.AssignVolume(ctx, request)
if err != nil {
glog.V(0).Infof("assign volume failure %v: %v", request, err)
return err
}
fileId, host = resp.FileId, resp.Url
return nil
}); err != nil {
return nil, fmt.Errorf("filerGrpcAddress assign volume: %v", err)
}
fileUrl := fmt.Sprintf("http://%s/%s", host, fileId)
bufReader := bytes.NewReader(buf)
uploadResult, err := operation.Upload(fileUrl, pages.f.Name, bufReader, false, "application/octet-stream", nil, "")
if err != nil {
glog.V(0).Infof("upload data %v to %s: %v", pages.f.Name, fileUrl, err)
return nil, fmt.Errorf("upload data: %v", err)
}
if uploadResult.Error != "" {
glog.V(0).Infof("upload failure %v to %s: %v", pages.f.Name, fileUrl, err)
return nil, fmt.Errorf("upload result: %v", uploadResult.Error)
}
return &filer_pb.FileChunk{
FileId: fileId,
Offset: offset,
Size: uint64(len(buf)),
Mtime: time.Now().UnixNano(),
ETag: uploadResult.ETag,
}, nil
}
func max(x, y int64) int64 {
if x > y {
return x
}
return y
}
|
