-
Notifications
You must be signed in to change notification settings - Fork 67
/
ads_processor.go
344 lines (306 loc) · 12.4 KB
/
ads_processor.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
/*
* Copyright The Kmesh 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.
*/
package ads
import (
"fmt"
config_cluster_v3 "github.com/envoyproxy/go-control-plane/envoy/config/cluster/v3"
config_endpoint_v3 "github.com/envoyproxy/go-control-plane/envoy/config/endpoint/v3"
config_listener_v3 "github.com/envoyproxy/go-control-plane/envoy/config/listener/v3"
config_route_v3 "github.com/envoyproxy/go-control-plane/envoy/config/route/v3"
service_discovery_v3 "github.com/envoyproxy/go-control-plane/envoy/service/discovery/v3"
resource_v3 "github.com/envoyproxy/go-control-plane/pkg/resource/v3"
"google.golang.org/protobuf/proto"
"google.golang.org/protobuf/types/known/anypb"
"istio.io/istio/pkg/slices"
"k8s.io/apimachinery/pkg/util/sets"
admin_v2 "kmesh.net/kmesh/api/v2/admin"
core_v2 "kmesh.net/kmesh/api/v2/core"
"kmesh.net/kmesh/pkg/constants"
"kmesh.net/kmesh/pkg/controller/config"
"kmesh.net/kmesh/pkg/utils/hash"
)
const (
apiVersionInfo = "v2"
)
type lastNonce struct {
cdsNonce string
edsNonce string
ldsNonce string
rdsNonce string
}
type processor struct {
Cache *AdsCache
ack *service_discovery_v3.DiscoveryRequest
req *service_discovery_v3.DiscoveryRequest
lastNonce *lastNonce
// the channel used to send domains to dns resolver. key is domain name and value is refreshrate
DnsResolverChan chan []*config_cluster_v3.Cluster
}
func newProcessor() *processor {
return &processor{
Cache: NewAdsCache(),
ack: nil,
req: nil,
lastNonce: &lastNonce{},
}
}
func newAdsRequest(typeUrl string, names []string, nonce string) *service_discovery_v3.DiscoveryRequest {
return &service_discovery_v3.DiscoveryRequest{
TypeUrl: typeUrl,
VersionInfo: "",
ResourceNames: names,
ResponseNonce: nonce,
ErrorDetail: nil,
Node: config.GetConfig(constants.KernelNativeMode).GetNode(),
}
}
func newAckRequest(resp *service_discovery_v3.DiscoveryResponse) *service_discovery_v3.DiscoveryRequest {
return &service_discovery_v3.DiscoveryRequest{
TypeUrl: resp.GetTypeUrl(),
VersionInfo: resp.GetVersionInfo(),
ResourceNames: []string{},
ResponseNonce: resp.GetNonce(),
ErrorDetail: nil,
Node: config.GetConfig(constants.KernelNativeMode).GetNode(),
}
}
// [Eventual consistency considerations](https://www.envoyproxy.io/docs/envoy/latest/api-docs/xds_protocol)
// In general, to avoid traffic drop, sequencing of updates should follow a make before break model, wherein:
// * CDS updates (if any) must always be pushed first.
// * EDS updates (if any) must arrive after CDS updates for the respective clusters.
// * LDS updates must arrive after corresponding CDS/EDS updates.
// * RDS updates related to the newly added listeners must arrive after CDS/EDS/LDS updates.
// * VHDS updates (if any) related to the newly added RouteConfigurations must arrive after RDS updates.
// * Stale CDS clusters and related EDS endpoints (ones no longer being referenced) can then be removed.
func (p *processor) processAdsResponse(resp *service_discovery_v3.DiscoveryResponse) {
var err error
log.Debugf("handle ads response, %#v\n", resp.GetTypeUrl())
p.ack = newAckRequest(resp)
if resp.GetResources() == nil {
return
}
switch resp.GetTypeUrl() {
case resource_v3.ClusterType:
err = p.handleCdsResponse(resp)
case resource_v3.EndpointType:
err = p.handleEdsResponse(resp)
case resource_v3.ListenerType:
err = p.handleLdsResponse(resp)
case resource_v3.RouteType:
err = p.handleRdsResponse(resp)
default:
err = fmt.Errorf("unsupported type url %s", resp.GetTypeUrl())
}
if err != nil {
log.Error(err)
}
}
func (p *processor) handleCdsResponse(resp *service_discovery_v3.DiscoveryResponse) error {
p.lastNonce.cdsNonce = resp.Nonce
current := sets.New[string]()
lastEdsClusterNames := p.Cache.edsClusterNames
p.Cache.edsClusterNames = nil
dnsClusters := []*config_cluster_v3.Cluster{}
for _, resource := range resp.GetResources() {
cluster := &config_cluster_v3.Cluster{}
if err := anypb.UnmarshalTo(resource, cluster, proto.UnmarshalOptions{}); err != nil {
log.Errorf("unmarshal cluster error: %v", err)
continue
}
current.Insert(cluster.GetName())
if cluster.GetType() == config_cluster_v3.Cluster_EDS {
p.Cache.edsClusterNames = append(p.Cache.edsClusterNames, cluster.GetName())
} else if cluster.GetType() == config_cluster_v3.Cluster_STRICT_DNS ||
cluster.GetType() == config_cluster_v3.Cluster_LOGICAL_DNS {
dnsClusters = append(dnsClusters, cluster)
}
// compare part[0] CDS now
// Cluster_EDS need compare tow parts, compare part[1] EDS in EDS handler
newHash := hash.Sum64String(resource.String())
if newHash != p.Cache.ClusterCache.GetCdsHash(cluster.GetName()) {
var status core_v2.ApiStatus
if cluster.GetType() == config_cluster_v3.Cluster_EDS {
status = core_v2.ApiStatus_WAITING
} else if cluster.GetType() == config_cluster_v3.Cluster_STRICT_DNS ||
cluster.GetType() == config_cluster_v3.Cluster_LOGICAL_DNS {
// dns typed cluster will be handled in dns module, skip update bpf map here
status = core_v2.ApiStatus_WAITING
} else {
status = core_v2.ApiStatus_UPDATE
}
log.Debugf("[CreateApiClusterByCds] update cluster %s, status %d, cluster.type %v",
cluster.GetName(), status, cluster.GetType())
p.Cache.ClusterCache.SetCdsHash(cluster.GetName(), newHash)
p.Cache.CreateApiClusterByCds(status, cluster)
} else {
log.Debugf("unchanged cluster %s", cluster.GetName())
}
}
// send dns clusters to dns resolver, even dnsClusters is empty, we need to send empty list to dns resolver to clear the cache
if p.DnsResolverChan != nil {
p.DnsResolverChan <- dnsClusters
}
removed := p.Cache.ClusterCache.GetResourceNames().Difference(current)
for key := range removed {
p.Cache.UpdateApiClusterStatus(key, core_v2.ApiStatus_DELETE)
}
if len(removed) > 0 {
log.Debugf("removed cluster: %v", removed.UnsortedList())
}
// Flush the clusters in these cases:
// 1. clusters need to be deleted
// 2. dns typed clusters update, we do not need to wait for eds update, because dns cluster has no eds following
// Note eds typed cluster, we do not flush to bpf map here, we need to wait for eds update.
p.Cache.ClusterCache.Flush()
if p.lastNonce.edsNonce == "" {
// initial subscribe to eds
p.req = newAdsRequest(resource_v3.EndpointType, p.Cache.edsClusterNames, "")
return nil
}
// when the list of eds typed clusters subscribed changed, we should resubscrbe to new eds.
if !slices.EqualUnordered(p.Cache.edsClusterNames, lastEdsClusterNames) {
// we cannot set the nonce here.
// There is a race: when xds server has pushed eds, but kmesh hasn't a chance to receive and process
// Then it will lead to this request been ignored, we will lose the new eds resource
p.req = newAdsRequest(resource_v3.EndpointType, p.Cache.edsClusterNames, "")
}
return nil
}
func (p *processor) handleEdsResponse(resp *service_discovery_v3.DiscoveryResponse) error {
var loadAssignment = &config_endpoint_v3.ClusterLoadAssignment{}
p.lastNonce.edsNonce = resp.Nonce
for _, resource := range resp.GetResources() {
if err := anypb.UnmarshalTo(resource, loadAssignment, proto.UnmarshalOptions{}); err != nil {
continue
}
cluster := p.Cache.ClusterCache.GetApiCluster(loadAssignment.GetClusterName())
// fix exceptional scenarios: receive eds push after cds has been deleted
if cluster == nil {
log.Debugf("cluster %s is deleted", loadAssignment.GetClusterName())
continue
}
apiStatus := cluster.ApiStatus
newHash := hash.Sum64String(resource.String())
// part[0] CDS is different or part[1] EDS is different
if apiStatus == core_v2.ApiStatus_WAITING ||
newHash != p.Cache.ClusterCache.GetEdsHash(loadAssignment.GetClusterName()) {
apiStatus = core_v2.ApiStatus_UPDATE
p.Cache.ClusterCache.SetEdsHash(loadAssignment.GetClusterName(), newHash)
log.Debugf("[CreateApiClusterByEds] update cluster %s", loadAssignment.GetClusterName())
p.Cache.CreateApiClusterByEds(apiStatus, loadAssignment)
} else {
log.Debugf("handleEdsResponse: unchanged cluster %s", loadAssignment.GetClusterName())
}
}
// EDS ack should contain all the eds cluster names, and since istiod can send partial eds to us, we use those set by handleCdsResponse
// Ad xds protocol spec, the non wildcard resource ack should contain all the names
p.ack.ResourceNames = p.Cache.edsClusterNames
if p.lastNonce.ldsNonce == "" {
// subscribe to lds only once per stream
p.req = newAdsRequest(resource_v3.ListenerType, nil, "")
}
p.Cache.ClusterCache.Flush()
return nil
}
func (p *processor) handleLdsResponse(resp *service_discovery_v3.DiscoveryResponse) error {
var (
err error
listener = &config_listener_v3.Listener{}
)
p.lastNonce.ldsNonce = resp.Nonce
current := sets.New[string]()
lastRouteNames := p.Cache.routeNames
p.Cache.routeNames = []string{}
for _, resource := range resp.GetResources() {
if err = anypb.UnmarshalTo(resource, listener, proto.UnmarshalOptions{}); err != nil {
continue
}
if listener.GetAddress() == nil {
// skip the listener without address
continue
}
current.Insert(listener.GetName())
apiStatus := core_v2.ApiStatus_UPDATE
newHash := hash.Sum64String(resource.String())
if newHash != p.Cache.ListenerCache.GetLdsHash(listener.GetName()) {
p.Cache.ListenerCache.AddOrUpdateLdsHash(listener.GetName(), newHash)
log.Debugf("[CreateApiListenerByLds] update %s", listener.GetName())
} else {
log.Debugf("[CreateApiListenerByLds] unchanged %s", listener.GetName())
apiStatus = core_v2.ApiStatus_UNCHANGED
}
p.Cache.CreateApiListenerByLds(apiStatus, listener)
}
removed := p.Cache.ListenerCache.GetResourceNames().Difference(current)
for key := range removed {
p.Cache.UpdateApiListenerStatus(key, core_v2.ApiStatus_DELETE)
}
p.Cache.ListenerCache.Flush()
if !slices.EqualUnordered(p.Cache.routeNames, lastRouteNames) {
// we cannot set the nonce here.
// There is a race: when xds server has pushed rds, but kmesh hasn't a chance to receive and process
// Then it will lead to this request been ignored, we will lose the new rds resource
p.req = newAdsRequest(resource_v3.RouteType, p.Cache.routeNames, "")
}
return nil
}
func (p *processor) handleRdsResponse(resp *service_discovery_v3.DiscoveryResponse) error {
routeConfiguration := &config_route_v3.RouteConfiguration{}
p.lastNonce.rdsNonce = resp.Nonce
current := sets.New[string]()
for _, resource := range resp.GetResources() {
if err := anypb.UnmarshalTo(resource, routeConfiguration, proto.UnmarshalOptions{}); err != nil {
continue
}
current.Insert(routeConfiguration.GetName())
newHash := hash.Sum64String(resource.String())
if newHash != p.Cache.RouteCache.GetRdsHash(routeConfiguration.GetName()) {
p.Cache.RouteCache.SetRdsHash(routeConfiguration.GetName(), newHash)
log.Debugf("[CreateApiRouteByRds] update %s", routeConfiguration.GetName())
p.Cache.CreateApiRouteByRds(core_v2.ApiStatus_UPDATE, routeConfiguration)
} else {
log.Debugf("[CreateApiRouteByRds] unchanged %s", routeConfiguration.GetName())
}
// if rds has no virtualhost, no need to subscribe this rds again in response
if routeConfiguration.GetVirtualHosts() != nil {
p.ack.ResourceNames = append(p.ack.ResourceNames, routeConfiguration.GetName())
}
}
removed := p.Cache.RouteCache.GetResourceNames().Difference(current)
for key := range removed {
p.Cache.RouteCache.UpdateApiRouteStatus(key, core_v2.ApiStatus_DELETE)
}
p.Cache.RouteCache.Flush()
return nil
}
func (p *processor) Reset() {
if p == nil {
return
}
p.lastNonce = &lastNonce{}
}
func ConfigResourcesIsEmpty(resources *admin_v2.ConfigResources) bool {
if resources == nil {
return true
}
count := len(resources.GetClusterConfigs()) + len(resources.GetRouteConfigs()) + len(resources.GetListenerConfigs())
return count == 0
}
func SetApiVersionInfo(resources *admin_v2.ConfigResources) {
if !ConfigResourcesIsEmpty(resources) {
resources.VersionInfo = apiVersionInfo
}
}