-
Notifications
You must be signed in to change notification settings - Fork 18.7k
/
sandbox_store.go
279 lines (236 loc) · 6.38 KB
/
sandbox_store.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
package libnetwork
import (
"encoding/json"
"sync"
"github.com/docker/docker/libnetwork/datastore"
"github.com/docker/docker/libnetwork/osl"
"github.com/sirupsen/logrus"
)
const (
sandboxPrefix = "sandbox"
)
type epState struct {
Eid string
Nid string
}
type sbState struct {
ID string
Cid string
c *Controller
dbIndex uint64
dbExists bool
Eps []epState
EpPriority map[string]int
// external servers have to be persisted so that on restart of a live-restore
// enabled daemon we get the external servers for the running containers.
//
// It is persisted as "ExtDNS2" for historical reasons. ExtDNS2 was used to
// handle migration between docker < 1.14 and >= 1.14. Before version 1.14 we
// used ExtDNS but with a []string. As it's unlikely that installations still
// have state from before 1.14, we've dropped the migration code.
ExtDNS []extDNSEntry `json:"ExtDNS2"`
}
func (sbs *sbState) Key() []string {
return []string{sandboxPrefix, sbs.ID}
}
func (sbs *sbState) KeyPrefix() []string {
return []string{sandboxPrefix}
}
func (sbs *sbState) Value() []byte {
b, err := json.Marshal(sbs)
if err != nil {
return nil
}
return b
}
func (sbs *sbState) SetValue(value []byte) error {
return json.Unmarshal(value, sbs)
}
func (sbs *sbState) Index() uint64 {
sb, err := sbs.c.SandboxByID(sbs.ID)
if err != nil {
return sbs.dbIndex
}
maxIndex := sb.dbIndex
if sbs.dbIndex > maxIndex {
maxIndex = sbs.dbIndex
}
return maxIndex
}
func (sbs *sbState) SetIndex(index uint64) {
sbs.dbIndex = index
sbs.dbExists = true
sb, err := sbs.c.SandboxByID(sbs.ID)
if err != nil {
return
}
sb.dbIndex = index
sb.dbExists = true
}
func (sbs *sbState) Exists() bool {
if sbs.dbExists {
return sbs.dbExists
}
sb, err := sbs.c.SandboxByID(sbs.ID)
if err != nil {
return false
}
return sb.dbExists
}
func (sbs *sbState) Skip() bool {
return false
}
func (sbs *sbState) New() datastore.KVObject {
return &sbState{c: sbs.c}
}
func (sbs *sbState) CopyTo(o datastore.KVObject) error {
dstSbs := o.(*sbState)
dstSbs.c = sbs.c
dstSbs.ID = sbs.ID
dstSbs.Cid = sbs.Cid
dstSbs.dbIndex = sbs.dbIndex
dstSbs.dbExists = sbs.dbExists
dstSbs.EpPriority = sbs.EpPriority
dstSbs.Eps = append(dstSbs.Eps, sbs.Eps...)
dstSbs.ExtDNS = append(dstSbs.ExtDNS, sbs.ExtDNS...)
return nil
}
func (sbs *sbState) DataScope() string {
return datastore.LocalScope
}
func (sb *Sandbox) storeUpdate() error {
sbs := &sbState{
c: sb.controller,
ID: sb.id,
Cid: sb.containerID,
EpPriority: sb.epPriority,
ExtDNS: sb.extDNS,
}
retry:
sbs.Eps = nil
for _, ep := range sb.Endpoints() {
// If the endpoint is not persisted then do not add it to
// the sandbox checkpoint
if ep.Skip() {
continue
}
eps := epState{
Nid: ep.getNetwork().ID(),
Eid: ep.ID(),
}
sbs.Eps = append(sbs.Eps, eps)
}
err := sb.controller.updateToStore(sbs)
if err == datastore.ErrKeyModified {
// When we get ErrKeyModified it is sufficient to just
// go back and retry. No need to get the object from
// the store because we always regenerate the store
// state from in memory sandbox state
goto retry
}
return err
}
func (sb *Sandbox) storeDelete() error {
sbs := &sbState{
c: sb.controller,
ID: sb.id,
Cid: sb.containerID,
dbIndex: sb.dbIndex,
dbExists: sb.dbExists,
}
return sb.controller.deleteFromStore(sbs)
}
func (c *Controller) sandboxCleanup(activeSandboxes map[string]interface{}) {
store := c.getStore()
if store == nil {
logrus.Error("Could not find local scope store while trying to cleanup sandboxes")
return
}
kvol, err := store.List(datastore.Key(sandboxPrefix), &sbState{c: c})
if err != nil && err != datastore.ErrKeyNotFound {
logrus.Errorf("failed to get sandboxes for scope %s: %v", store.Scope(), err)
return
}
// It's normal for no sandboxes to be found. Just bail out.
if err == datastore.ErrKeyNotFound {
return
}
for _, kvo := range kvol {
sbs := kvo.(*sbState)
sb := &Sandbox{
id: sbs.ID,
controller: sbs.c,
containerID: sbs.Cid,
extDNS: sbs.ExtDNS,
endpoints: []*Endpoint{},
populatedEndpoints: map[string]struct{}{},
dbIndex: sbs.dbIndex,
isStub: true,
dbExists: true,
}
msg := " for cleanup"
create := true
isRestore := false
if val, ok := activeSandboxes[sb.ID()]; ok {
msg = ""
sb.isStub = false
isRestore = true
opts := val.([]SandboxOption)
sb.processOptions(opts...)
sb.restorePath()
create = !sb.config.useDefaultSandBox
}
sb.osSbox, err = osl.NewSandbox(sb.Key(), create, isRestore)
if err != nil {
logrus.Errorf("failed to create osl sandbox while trying to restore sandbox %.7s%s: %v", sb.ID(), msg, err)
continue
}
c.mu.Lock()
c.sandboxes[sb.id] = sb
c.mu.Unlock()
for _, eps := range sbs.Eps {
n, err := c.getNetworkFromStore(eps.Nid)
var ep *Endpoint
if err != nil {
logrus.Errorf("getNetworkFromStore for nid %s failed while trying to build sandbox for cleanup: %v", eps.Nid, err)
n = &network{id: eps.Nid, ctrlr: c, drvOnce: &sync.Once{}, persist: true}
ep = &Endpoint{id: eps.Eid, network: n, sandboxID: sbs.ID}
} else {
ep, err = n.getEndpointFromStore(eps.Eid)
if err != nil {
logrus.Errorf("getEndpointFromStore for eid %s failed while trying to build sandbox for cleanup: %v", eps.Eid, err)
ep = &Endpoint{id: eps.Eid, network: n, sandboxID: sbs.ID}
}
}
if _, ok := activeSandboxes[sb.ID()]; ok && err != nil {
logrus.Errorf("failed to restore endpoint %s in %s for container %s due to %v", eps.Eid, eps.Nid, sb.ContainerID(), err)
continue
}
sb.addEndpoint(ep)
}
if _, ok := activeSandboxes[sb.ID()]; !ok {
logrus.Infof("Removing stale sandbox %s (%s)", sb.id, sb.containerID)
if err := sb.delete(true); err != nil {
logrus.Errorf("Failed to delete sandbox %s while trying to cleanup: %v", sb.id, err)
}
continue
}
// reconstruct osl sandbox field
if !sb.config.useDefaultSandBox {
if err := sb.restoreOslSandbox(); err != nil {
logrus.Errorf("failed to populate fields for osl sandbox %s", sb.ID())
continue
}
} else {
c.sboxOnce.Do(func() {
c.defOsSbox = sb.osSbox
})
}
for _, ep := range sb.endpoints {
// Watch for service records
if !c.isAgent() {
c.watchSvcRecord(ep)
}
}
}
}