/
migrate.go
247 lines (200 loc) · 6.27 KB
/
migrate.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
// Package migration provides the primitives for migration in LXD.
//
// See https://github.com/canonical/lxd/blob/main/doc/migration.md for a complete
// description.
package main
import (
"context"
"fmt"
"net/http"
"sync"
"time"
"github.com/gorilla/websocket"
"google.golang.org/protobuf/proto"
"github.com/canonical/lxd/lxd/idmap"
"github.com/canonical/lxd/lxd/instance"
"github.com/canonical/lxd/lxd/migration"
"github.com/canonical/lxd/lxd/operations"
"github.com/canonical/lxd/shared"
"github.com/canonical/lxd/shared/api"
)
type migrationFields struct {
controlLock sync.Mutex
conns map[string]*migrationConn
// container specific fields
live bool
instanceOnly bool
instance instance.Instance
// storage specific fields
volumeOnly bool
allowInconsistent bool
}
func (c *migrationFields) send(m proto.Message) error {
/* gorilla websocket doesn't allow concurrent writes, and
* panic()s if it sees them (which is reasonable). If e.g. we
* happen to fail, get scheduled, start our write, then get
* unscheduled before the write is bit to a new thread which is
* receiving an error from the other side (due to our previous
* close), we can engage in these concurrent writes, which
* casuses the whole daemon to panic.
*
* Instead, let's lock sends to the controlConn so that we only ever
* write one message at the time.
*/
c.controlLock.Lock()
defer c.controlLock.Unlock()
conn, err := c.conns[api.SecretNameControl].WebSocket(context.TODO())
if err != nil {
return fmt.Errorf("Control connection not initialized: %w", err)
}
_ = conn.SetWriteDeadline(time.Now().Add(time.Second * 30))
err = migration.ProtoSend(conn, m)
if err != nil {
return err
}
return nil
}
func (c *migrationFields) recv(m proto.Message) error {
conn, err := c.conns[api.SecretNameControl].WebSocket(context.TODO())
if err != nil {
return fmt.Errorf("Control connection not initialized: %w", err)
}
return migration.ProtoRecv(conn, m)
}
func (c *migrationFields) disconnect() {
c.controlLock.Lock()
ctx, cancel := context.WithTimeout(context.TODO(), time.Second)
defer cancel()
conn, _ := c.conns[api.SecretNameControl].WebSocket(ctx)
if conn != nil {
closeMsg := websocket.FormatCloseMessage(websocket.CloseNormalClosure, "")
_ = conn.SetWriteDeadline(time.Now().Add(time.Second * 30))
_ = conn.WriteMessage(websocket.CloseMessage, closeMsg)
}
c.controlLock.Unlock()
/* Below we just Close(), which doesn't actually write to the
* websocket, it just closes the underlying connection. If e.g. there
* is still a filesystem transfer going on, but the other side has run
* out of disk space, writing an actual CloseMessage here will cause
* gorilla websocket to panic. Instead, we just force close this
* connection, since we report the error over the control channel
* anyway.
*/
for _, conn := range c.conns {
conn.Close()
}
}
func (c *migrationFields) sendControl(err error) {
c.controlLock.Lock()
conn, _ := c.conns[api.SecretNameControl].WebSocket(context.TODO())
if conn != nil {
migration.ProtoSendControl(conn, err)
}
c.controlLock.Unlock()
if err != nil {
c.disconnect()
}
}
func (c *migrationFields) controlChannel() <-chan *migration.ControlResponse {
ch := make(chan *migration.ControlResponse)
go func() {
resp := migration.ControlResponse{}
err := c.recv(&resp.MigrationControl)
if err != nil {
resp.Err = err
ch <- &resp
return
}
ch <- &resp
}()
return ch
}
type migrationSourceWs struct {
migrationFields
clusterMoveSourceName string
pushCertificate string
pushOperationURL string
pushSecrets map[string]string
}
func (s *migrationSourceWs) Metadata() any {
secrets := make(shared.Jmap, len(s.conns))
for connName, conn := range s.conns {
secrets[connName] = conn.Secret()
}
return secrets
}
func (s *migrationSourceWs) Connect(op *operations.Operation, r *http.Request, w http.ResponseWriter) error {
incomingSecret := r.FormValue("secret")
if incomingSecret == "" {
return api.StatusErrorf(http.StatusBadRequest, "Missing migration source secret")
}
for connName, conn := range s.conns {
if incomingSecret != conn.Secret() {
continue
}
err := conn.AcceptIncoming(r, w)
if err != nil {
return fmt.Errorf("Failed accepting incoming migration source %q connection: %w", connName, err)
}
return nil
}
// If we didn't find the right secret, the user provided a bad one, so return 403, not 404, since this
// operation actually exists.
return api.StatusErrorf(http.StatusForbidden, "Invalid migration source secret")
}
type migrationSink struct {
migrationFields
url string
push bool
clusterMoveSourceName string
refresh bool
}
// MigrationSinkArgs arguments to configure migration sink.
type migrationSinkArgs struct {
// General migration fields
Dialer *websocket.Dialer
Push bool
Secrets map[string]string
URL string
// Instance specific fields
Instance instance.Instance
InstanceOnly bool
Idmap *idmap.IdmapSet
Live bool
Refresh bool
ClusterMoveSourceName string
Snapshots []*migration.Snapshot
// Storage specific fields
VolumeOnly bool
VolumeSize int64
// Transport specific fields
RsyncFeatures []string
}
// Metadata returns metadata for the migration sink.
func (s *migrationSink) Metadata() any {
secrets := make(shared.Jmap, len(s.conns))
for connName, conn := range s.conns {
secrets[connName] = conn.Secret()
}
return secrets
}
// Connect connects to the migration source.
func (s *migrationSink) Connect(op *operations.Operation, r *http.Request, w http.ResponseWriter) error {
incomingSecret := r.FormValue("secret")
if incomingSecret == "" {
return api.StatusErrorf(http.StatusBadRequest, "Missing migration sink secret")
}
for connName, conn := range s.conns {
if incomingSecret != conn.Secret() {
continue
}
err := conn.AcceptIncoming(r, w)
if err != nil {
return fmt.Errorf("Failed accepting incoming migration sink %q connection: %w", connName, err)
}
return nil
}
// If we didn't find the right secret, the user provided a bad one, so return 403, not 404, since this
// operation actually exists.
return api.StatusErrorf(http.StatusForbidden, "Invalid migration sink secret")
}