forked from metallb/metallb
/
bgp_controller.go
380 lines (349 loc) · 11 KB
/
bgp_controller.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
// Copyright 2017 Google Inc.
//
// 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 main
import (
"fmt"
"net"
"reflect"
"sort"
"strconv"
"go.universe.tf/metallb/internal/bgp"
bgpfrr "go.universe.tf/metallb/internal/bgp/frr"
bgpnative "go.universe.tf/metallb/internal/bgp/native"
"go.universe.tf/metallb/internal/config"
"go.universe.tf/metallb/internal/k8s/epslices"
k8snodes "go.universe.tf/metallb/internal/k8s/nodes"
"go.universe.tf/metallb/internal/logging"
v1 "k8s.io/api/core/v1"
"k8s.io/apimachinery/pkg/labels"
"github.com/go-kit/log"
"github.com/go-kit/log/level"
"github.com/pkg/errors"
)
type bgpImplementation string
const (
bgpNative bgpImplementation = "native"
bgpFrr bgpImplementation = "frr"
)
type peer struct {
cfg *config.Peer
session bgp.Session
}
type bgpController struct {
logger log.Logger
myNode string
nodeLabels labels.Set
peers []*peer
svcAds map[string][]*bgp.Advertisement
bgpType bgpImplementation
sessionManager bgp.SessionManager
}
func (c *bgpController) SetConfig(l log.Logger, cfg *config.Config) error {
newPeers := make([]*peer, 0, len(cfg.Peers))
newPeers:
for _, p := range cfg.Peers {
for i, ep := range c.peers {
if ep == nil {
continue
}
if reflect.DeepEqual(p, ep.cfg) {
newPeers = append(newPeers, ep)
c.peers[i] = nil
continue newPeers
}
}
// No existing peers match, create a new one.
newPeers = append(newPeers, &peer{
cfg: p,
})
}
oldPeers := c.peers
c.peers = newPeers
for _, p := range oldPeers {
if p == nil {
continue
}
level.Info(l).Log("event", "peerRemoved", "peer", p.cfg.Addr, "reason", "removedFromConfig", "msg", "peer deconfigured, closing BGP session")
if p.session != nil {
if err := p.session.Close(); err != nil {
level.Error(l).Log("op", "setConfig", "error", err, "peer", p.cfg.Addr, "msg", "failed to shut down BGP session")
}
}
level.Debug(l).Log("event", "peerRemoved", "peer", p.cfg.Addr, "reason", "removedFromConfig", "msg", "peer deconfigured, BGP session closed")
}
err := c.syncBFDProfiles(cfg.BFDProfiles)
if err != nil {
return errors.Wrap(err, "failed to sync bfd profiles")
}
err = c.sessionManager.SyncExtraInfo(cfg.BGPExtras)
if err != nil {
return errors.Wrap(err, "failed to sync extra info")
}
return c.syncPeers(l)
}
// hasHealthyEndpoint return true if this node has at least one healthy endpoint.
// It only checks nodes matching the given filterNode function.
func hasHealthyEndpoint(eps epslices.EpsOrSlices, filterNode func(*string) bool) bool {
ready := map[string]bool{}
switch eps.Type {
case epslices.Eps:
for _, subset := range eps.EpVal.Subsets {
for _, ep := range subset.Addresses {
if filterNode(ep.NodeName) {
continue
}
if _, ok := ready[ep.IP]; !ok {
// Only set true if nothing else has expressed an
// opinion. This means that false will take precedence
// if there's any unready ports for a given endpoint.
ready[ep.IP] = true
}
}
for _, ep := range subset.NotReadyAddresses {
ready[ep.IP] = false
}
}
case epslices.Slices:
for _, slice := range eps.SlicesVal {
for _, ep := range slice.Endpoints {
node := ep.NodeName
if filterNode(node) {
continue
}
for _, addr := range ep.Addresses {
if _, ok := ready[addr]; !ok && epslices.IsConditionServing(ep.Conditions) {
// Only set true if nothing else has expressed an
// opinion. This means that false will take precedence
// if there's any unready ports for a given endpoint.
ready[addr] = true
}
if !epslices.IsConditionServing(ep.Conditions) {
ready[addr] = false
}
}
}
}
}
for _, r := range ready {
if r {
// At least one fully healthy endpoint on this machine.
return true
}
}
return false
}
func (c *bgpController) ShouldAnnounce(l log.Logger, name string, _ []net.IP, pool *config.Pool, svc *v1.Service, eps epslices.EpsOrSlices, nodes map[string]*v1.Node) string {
if !poolMatchesNodeBGP(pool, c.myNode) {
level.Debug(l).Log("event", "skipping should announce bgp", "service", name, "reason", "pool not matching my node")
return "notOwner"
}
if k8snodes.IsNetworkUnavailable(nodes[c.myNode]) {
level.Debug(l).Log("event", "skipping should announce bgp", "service", name, "reason", "speaker's node has NodeNetworkUnavailable condition")
return "nodeNetworkUnavailable"
}
// Should we advertise?
// Yes, if externalTrafficPolicy is
// Cluster && any healthy endpoint exists
// or
// Local && there's a ready local endpoint.
filterNode := func(toFilter *string) bool {
if toFilter == nil || *toFilter != c.myNode {
return true
}
return false
}
if svc.Spec.ExternalTrafficPolicy == v1.ServiceExternalTrafficPolicyTypeLocal && !hasHealthyEndpoint(eps, filterNode) {
return "noLocalEndpoints"
} else if !hasHealthyEndpoint(eps, func(toFilter *string) bool { return false }) {
return "noEndpoints"
}
return ""
}
// Called when either the peer list or node labels have changed,
// implying that the set of running BGP sessions may need tweaking.
func (c *bgpController) syncPeers(l log.Logger) error {
var (
errs int
needUpdateAds bool
)
for _, p := range c.peers {
// First, determine if the peering should be active for this
// node.
shouldRun := false
if len(p.cfg.NodeSelectors) == 0 {
shouldRun = true
}
for _, ns := range p.cfg.NodeSelectors {
if ns.Matches(c.nodeLabels) {
shouldRun = true
break
}
}
// Now, compare current state to intended state, and correct.
if p.session != nil && !shouldRun {
// Oops, session is running but shouldn't be. Shut it down.
level.Info(l).Log("event", "peerRemoved", "peer", p.cfg.Addr, "reason", "filteredByNodeSelector", "msg", "peer deconfigured, closing BGP session")
if err := p.session.Close(); err != nil {
level.Error(l).Log("op", "syncPeers", "error", err, "peer", p.cfg.Addr, "msg", "failed to shut down BGP session")
}
p.session = nil
} else if p.session == nil && shouldRun {
// Session doesn't exist, but should be running. Create
// it.
level.Info(l).Log("event", "peerAdded", "peer", p.cfg.Addr, "msg", "peer configured, starting BGP session")
var routerID net.IP
if p.cfg.RouterID != nil {
routerID = p.cfg.RouterID
}
s, err := c.sessionManager.NewSession(c.logger,
bgp.SessionParameters{
PeerAddress: net.JoinHostPort(p.cfg.Addr.String(), strconv.Itoa(int(p.cfg.Port))),
SourceAddress: p.cfg.SrcAddr,
MyASN: p.cfg.MyASN,
RouterID: routerID,
PeerASN: p.cfg.ASN,
HoldTime: p.cfg.HoldTime,
KeepAliveTime: p.cfg.KeepaliveTime,
Password: p.cfg.Password,
CurrentNode: c.myNode,
BFDProfile: p.cfg.BFDProfile,
EBGPMultiHop: p.cfg.EBGPMultiHop,
SessionName: p.cfg.Name,
VRFName: p.cfg.VRF,
},
)
if err != nil {
level.Error(l).Log("op", "syncPeers", "error", err, "peer", p.cfg.Addr, "msg", "failed to create BGP session")
errs++
} else {
p.session = s
needUpdateAds = true
}
}
}
if needUpdateAds {
// Some new sessions came up, resync advertisement state.
if err := c.updateAds(); err != nil {
level.Error(l).Log("op", "updateAds", "error", err, "msg", "failed to update BGP advertisements")
return err
}
}
if errs > 0 {
return fmt.Errorf("%d BGP sessions failed to start", errs)
}
return nil
}
func (c *bgpController) syncBFDProfiles(profiles map[string]*config.BFDProfile) error {
return c.sessionManager.SyncBFDProfiles(profiles)
}
func (c *bgpController) SetBalancer(l log.Logger, name string, lbIPs []net.IP, pool *config.Pool, _ service, _ *v1.Service) error {
c.svcAds[name] = nil
for _, lbIP := range lbIPs {
for _, adCfg := range pool.BGPAdvertisements {
// skipping if this node is not enabled for this advertisement
if !adCfg.Nodes[c.myNode] {
continue
}
m := net.CIDRMask(adCfg.AggregationLength, 32)
if lbIP.To4() == nil {
m = net.CIDRMask(adCfg.AggregationLengthV6, 128)
}
ad := &bgp.Advertisement{
Prefix: &net.IPNet{
IP: lbIP.Mask(m),
Mask: m,
},
LocalPref: adCfg.LocalPref,
}
if len(adCfg.Peers) > 0 {
ad.Peers = make([]string, 0, len(adCfg.Peers))
ad.Peers = append(ad.Peers, adCfg.Peers...)
}
for comm := range adCfg.Communities {
ad.Communities = append(ad.Communities, comm)
}
sort.Slice(ad.Communities, func(i, j int) bool { return ad.Communities[i].LessThan(ad.Communities[j]) })
c.svcAds[name] = append(c.svcAds[name], ad)
}
}
if err := c.updateAds(); err != nil {
return err
}
level.Info(l).Log("event", "updatedAdvertisements", "numAds", len(c.svcAds[name]), "msg", "making advertisements using BGP")
return nil
}
func (c *bgpController) updateAds() error {
var allAds []*bgp.Advertisement
for _, ads := range c.svcAds {
// This list might contain duplicates, but that's fine,
// they'll get compacted by the session code when it's
// calculating advertisements.
//
// TODO: be more intelligent about compacting advertisements
// and detecting conflicting advertisements.
allAds = append(allAds, ads...)
}
for _, peer := range c.peers {
if peer.session == nil {
continue
}
if err := peer.session.Set(allAds...); err != nil {
return err
}
}
return nil
}
func (c *bgpController) DeleteBalancer(l log.Logger, name, reason string) error {
if _, ok := c.svcAds[name]; !ok {
return nil
}
delete(c.svcAds, name)
return c.updateAds()
}
func (c *bgpController) SetNode(l log.Logger, node *v1.Node) error {
if c.myNode != node.Name {
return nil
}
nodeLabels := node.Labels
if nodeLabels == nil {
nodeLabels = map[string]string{}
}
ns := labels.Set(nodeLabels)
if c.nodeLabels != nil && labels.Equals(c.nodeLabels, ns) {
// Node labels unchanged, no action required.
return nil
}
c.nodeLabels = ns
level.Info(l).Log("event", "nodeLabelsChanged", "msg", "Node labels changed, resyncing BGP peers")
return c.syncPeers(l)
}
// Create a new 'bgp.SessionManager' of type 'bgpType'.
var newBGP = func(bgpType bgpImplementation, l log.Logger, logLevel logging.Level) bgp.SessionManager {
switch bgpType {
case bgpNative:
return bgpnative.NewSessionManager(l)
case bgpFrr:
return bgpfrr.NewSessionManager(l, logLevel)
default:
panic(fmt.Sprintf("unsupported BGP implementation type: %s", bgpType))
}
}
func poolMatchesNodeBGP(pool *config.Pool, node string) bool {
for _, adv := range pool.BGPAdvertisements {
if adv.Nodes[node] {
return true
}
}
return false
}