/
node.go
839 lines (767 loc) · 28.6 KB
/
node.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
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
package node
import (
"errors"
"fmt"
"io/ioutil"
"net"
"os"
"path/filepath"
"strconv"
"strings"
"sync"
"time"
kapi "k8s.io/api/core/v1"
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
"k8s.io/apimachinery/pkg/labels"
"k8s.io/apimachinery/pkg/selection"
"k8s.io/apimachinery/pkg/util/wait"
"k8s.io/client-go/informers"
"k8s.io/client-go/kubernetes"
clientset "k8s.io/client-go/kubernetes"
"k8s.io/client-go/tools/cache"
"k8s.io/client-go/tools/record"
"k8s.io/client-go/util/retry"
"k8s.io/klog/v2"
utilnet "k8s.io/utils/net"
honode "github.com/ovn-org/ovn-kubernetes/go-controller/hybrid-overlay/pkg/controller"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/cni"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/config"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/factory"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/informer"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/kube"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/node/controllers/upgrade"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/types"
"github.com/ovn-org/ovn-kubernetes/go-controller/pkg/util"
"github.com/vishvananda/netlink"
)
// OvnNode is the object holder for utilities meant for node management
type OvnNode struct {
name string
client clientset.Interface
Kube kube.Interface
watchFactory factory.NodeWatchFactory
stopChan chan struct{}
recorder record.EventRecorder
gateway Gateway
}
// NewNode creates a new controller for node management
func NewNode(kubeClient kubernetes.Interface, wf factory.NodeWatchFactory, name string, stopChan chan struct{}, eventRecorder record.EventRecorder) *OvnNode {
return &OvnNode{
name: name,
client: kubeClient,
Kube: &kube.Kube{KClient: kubeClient},
watchFactory: wf,
stopChan: stopChan,
recorder: eventRecorder,
}
}
func clearOVSFlowTargets() error {
_, _, err := util.RunOVSVsctl(
"--",
"clear", "bridge", "br-int", "netflow",
"--",
"clear", "bridge", "br-int", "sflow",
"--",
"clear", "bridge", "br-int", "ipfix",
)
if err != nil {
return err
}
return nil
}
// collectorsString joins all HostPort entry into a string that is acceptable as
// target by the ovs-vsctl command. If an entry has an empty host, it uses the Node IP
func collectorsString(node *kapi.Node, targets []config.HostPort) (string, error) {
if len(targets) == 0 {
return "", errors.New("collector targets can't be empty")
}
var joined strings.Builder
for n, v := range targets {
if n == 0 {
joined.WriteByte('"')
} else {
joined.WriteString(`","`)
}
var host string
if v.Host != nil && len(*v.Host) != 0 {
host = v.Host.String()
} else {
var err error
if host, err = util.GetNodePrimaryIP(node); err != nil {
return "", fmt.Errorf("composing flow collectors' IPs: %w", err)
}
}
joined.WriteString(util.JoinHostPortInt32(host, v.Port))
}
joined.WriteByte('"')
return joined.String(), nil
}
func setOVSFlowTargets(node *kapi.Node) error {
if len(config.Monitoring.NetFlowTargets) != 0 {
collectors, err := collectorsString(node, config.Monitoring.NetFlowTargets)
if err != nil {
return fmt.Errorf("error joining NetFlow targets: %w", err)
}
_, stderr, err := util.RunOVSVsctl(
"--",
"--id=@netflow",
"create",
"netflow",
fmt.Sprintf("targets=[%s]", collectors),
"active_timeout=60",
"--",
"set", "bridge", "br-int", "netflow=@netflow",
)
if err != nil {
return fmt.Errorf("error setting NetFlow: %v\n %q", err, stderr)
}
}
if len(config.Monitoring.SFlowTargets) != 0 {
collectors, err := collectorsString(node, config.Monitoring.SFlowTargets)
if err != nil {
return fmt.Errorf("error joining SFlow targets: %w", err)
}
_, stderr, err := util.RunOVSVsctl(
"--",
"--id=@sflow",
"create",
"sflow",
"agent="+types.SFlowAgent,
fmt.Sprintf("targets=[%s]", collectors),
"--",
"set", "bridge", "br-int", "sflow=@sflow",
)
if err != nil {
return fmt.Errorf("error setting SFlow: %v\n %q", err, stderr)
}
}
if len(config.Monitoring.IPFIXTargets) != 0 {
collectors, err := collectorsString(node, config.Monitoring.IPFIXTargets)
if err != nil {
return fmt.Errorf("error joining IPFIX targets: %w", err)
}
args := []string{
"--",
"--id=@ipfix",
"create",
"ipfix",
fmt.Sprintf("targets=[%s]", collectors),
fmt.Sprintf("cache_active_timeout=%d", config.IPFIX.CacheActiveTimeout),
}
if config.IPFIX.CacheMaxFlows != 0 {
args = append(args, fmt.Sprintf("cache_max_flows=%d", config.IPFIX.CacheMaxFlows))
}
if config.IPFIX.Sampling != 0 {
args = append(args, fmt.Sprintf("sampling=%d", config.IPFIX.Sampling))
}
args = append(args, "--", "set", "bridge", "br-int", "ipfix=@ipfix")
_, stderr, err := util.RunOVSVsctl(args...)
if err != nil {
return fmt.Errorf("error setting IPFIX: %v\n %q", err, stderr)
}
}
return nil
}
func setupOVNNode(node *kapi.Node) error {
var err error
encapIP := config.Default.EncapIP
if encapIP == "" {
encapIP, err = util.GetNodePrimaryIP(node)
if err != nil {
return fmt.Errorf("failed to obtain local IP from node %q: %v", node.Name, err)
}
config.Default.EncapIP = encapIP
} else {
if ip := net.ParseIP(encapIP); ip == nil {
return fmt.Errorf("invalid encapsulation IP provided %q", encapIP)
}
}
setExternalIdsCmd := []string{
"set",
"Open_vSwitch",
".",
fmt.Sprintf("external_ids:ovn-encap-type=%s", config.Default.EncapType),
fmt.Sprintf("external_ids:ovn-encap-ip=%s", encapIP),
fmt.Sprintf("external_ids:ovn-remote-probe-interval=%d",
config.Default.InactivityProbe),
fmt.Sprintf("external_ids:ovn-openflow-probe-interval=%d",
config.Default.OpenFlowProbe),
// bundle-idle-timeout default value is 10s, it should be set
// as high as the ovn-openflow-probe-interval to allow ovn-controller
// to finish computation specially with complex acl configuration with port range.
fmt.Sprintf("other_config:bundle-idle-timeout=%d",
config.Default.OpenFlowProbe),
fmt.Sprintf("external_ids:hostname=\"%s\"", node.Name),
fmt.Sprintf("external_ids:ovn-monitor-all=%t", config.Default.MonitorAll),
fmt.Sprintf("external_ids:ovn-enable-lflow-cache=%t", config.Default.LFlowCacheEnable),
}
if config.Default.LFlowCacheLimit > 0 {
setExternalIdsCmd = append(setExternalIdsCmd,
fmt.Sprintf("external_ids:ovn-limit-lflow-cache=%d", config.Default.LFlowCacheLimit),
)
}
if config.Default.LFlowCacheLimitKb > 0 {
setExternalIdsCmd = append(setExternalIdsCmd,
fmt.Sprintf("external_ids:ovn-limit-lflow-cache-kb=%d", config.Default.LFlowCacheLimitKb),
)
}
_, stderr, err := util.RunOVSVsctl(setExternalIdsCmd...)
if err != nil {
return fmt.Errorf("error setting OVS external IDs: %v\n %q", err, stderr)
}
// If EncapPort is not the default tell sbdb to use specified port.
if config.Default.EncapPort != config.DefaultEncapPort {
systemID, err := util.GetNodeChassisID()
if err != nil {
return err
}
uuid, _, err := util.RunOVNSbctl("--data=bare", "--no-heading", "--columns=_uuid", "find", "Encap",
fmt.Sprintf("chassis_name=%s", systemID))
if err != nil {
return err
}
if len(uuid) == 0 {
return fmt.Errorf("unable to find encap uuid to set geneve port for chassis %s", systemID)
}
_, stderr, errSet := util.RunOVNSbctl("set", "encap", uuid,
fmt.Sprintf("options:dst_port=%d", config.Default.EncapPort),
)
if errSet != nil {
return fmt.Errorf("error setting OVS encap-port: %v\n %q", errSet, stderr)
}
}
// clear stale ovs flow targets if needed
err = clearOVSFlowTargets()
if err != nil {
return fmt.Errorf("error clearing stale ovs flow targets: %q", err)
}
// set new ovs flow targets if needed
err = setOVSFlowTargets(node)
if err != nil {
return fmt.Errorf("error setting ovs flow targets: %q", err)
}
return nil
}
func isOVNControllerReady() (bool, error) {
// check node's connection status
runDir := util.GetOvnRunDir()
pid, err := ioutil.ReadFile(runDir + "ovn-controller.pid")
if err != nil {
return false, fmt.Errorf("unknown pid for ovn-controller process: %v", err)
}
ctlFile := runDir + fmt.Sprintf("ovn-controller.%s.ctl", strings.TrimSuffix(string(pid), "\n"))
ret, _, err := util.RunOVSAppctl("-t", ctlFile, "connection-status")
if err != nil {
return false, fmt.Errorf("could not get connection status: %w", err)
}
klog.Infof("Node connection status = %s", ret)
if ret != "connected" {
return false, nil
}
// check whether br-int exists on node
_, _, err = util.RunOVSVsctl("--", "br-exists", "br-int")
if err != nil {
return false, nil
}
// check by dumping br-int flow entries
stdout, _, err := util.RunOVSOfctl("dump-aggregate", "br-int")
if err != nil {
klog.V(5).Infof("Error dumping aggregate flows: %v", err)
return false, nil
}
hasFlowCountZero := strings.Contains(stdout, "flow_count=0")
if hasFlowCountZero {
klog.V(5).Info("Got a flow count of 0 when dumping flows for node")
return false, nil
}
return true, nil
}
// Starting with v21.03.0 OVN sets OVS.Interface.external-id:ovn-installed
// and OVNSB.Port_Binding.up when all OVS flows associated to a
// logical port have been successfully programmed.
// OVS.Interface.external-id:ovn-installed can only be used correctly
// in a combination with OVS.Interface.external-id:iface-id-ver
func getOVNIfUpCheckMode() (bool, error) {
if config.OvnKubeNode.DisableOVNIfaceIdVer {
klog.Infof("'iface-id-ver' is manually disabled, ovn-installed feature can't be used")
return false, nil
}
if _, stderr, err := util.RunOVNSbctl("--columns=up", "list", "Port_Binding"); err != nil {
if strings.Contains(stderr, "does not contain a column") {
klog.Infof("Falling back to using legacy OVS flow readiness checks")
return false, nil
}
return false, fmt.Errorf("failed to check if port_binding is supported in OVN, stderr: %q, error: %v",
stderr, err)
}
klog.Infof("Detected support for port binding with external IDs")
return true, nil
}
// Start learns the subnets assigned to it by the master controller
// and calls the SetupNode script which establishes the logical switch
func (n *OvnNode) Start(wg *sync.WaitGroup) error {
var err error
var node *kapi.Node
var subnets []*net.IPNet
var mgmtPort ManagementPort
var mgmtPortConfig *managementPortConfig
var cniServer *cni.Server
var isOvnUpEnabled bool
klog.Infof("OVN Kube Node initialization, Mode: %s", config.OvnKubeNode.Mode)
// Setting debug log level during node bring up to expose bring up process.
// Log level is returned to configured value when bring up is complete.
var level klog.Level
if err := level.Set("5"); err != nil {
klog.Errorf("Setting klog \"loglevel\" to 5 failed, err: %v", err)
}
// Start and sync the watch factory to begin listening for events
if err := n.watchFactory.Start(); err != nil {
return err
}
if node, err = n.Kube.GetNode(n.name); err != nil {
return fmt.Errorf("error retrieving node %s: %v", n.name, err)
}
nodeAddrStr, err := util.GetNodePrimaryIP(node)
if err != nil {
return err
}
nodeAddr := net.ParseIP(nodeAddrStr)
if nodeAddr == nil {
return fmt.Errorf("failed to parse kubernetes node IP address. %v", err)
}
if config.OvnKubeNode.Mode != types.NodeModeDPUHost {
for _, auth := range []config.OvnAuthConfig{config.OvnNorth, config.OvnSouth} {
if err := auth.SetDBAuth(); err != nil {
return err
}
}
err = setupOVNNode(node)
if err != nil {
return err
}
}
// First wait for the node logical switch to be created by the Master, timeout is 300s.
err = wait.PollImmediate(500*time.Millisecond, 300*time.Second, func() (bool, error) {
if node, err = n.Kube.GetNode(n.name); err != nil {
klog.Infof("Waiting to retrieve node %s: %v", n.name, err)
return false, nil
}
subnets, err = util.ParseNodeHostSubnetAnnotation(node)
if err != nil {
klog.Infof("Waiting for node %s to start, no annotation found on node for subnet: %v", n.name, err)
return false, nil
}
return true, nil
})
if err != nil {
return fmt.Errorf("timed out waiting for node's: %q logical switch: %v", n.name, err)
}
klog.Infof("Node %s ready for ovn initialization with subnet %s", n.name, util.JoinIPNets(subnets, ","))
if config.OvnKubeNode.Mode != types.NodeModeDPUHost {
isOvnUpEnabled, err = getOVNIfUpCheckMode()
if err != nil {
return err
}
}
// Create CNI Server
if config.OvnKubeNode.Mode != types.NodeModeDPU {
kclient, ok := n.Kube.(*kube.Kube)
if !ok {
return fmt.Errorf("cannot get kubeclient for starting CNI server")
}
cniServer, err = cni.NewCNIServer("", isOvnUpEnabled, n.watchFactory, kclient.KClient)
if err != nil {
return err
}
}
// Setup Management port and gateway
mgmtPort = NewManagementPort(n.name, subnets)
nodeAnnotator := kube.NewNodeAnnotator(n.Kube, node.Name)
waiter := newStartupWaiter()
mgmtPortConfig, err = mgmtPort.Create(nodeAnnotator, waiter)
if err != nil {
return err
}
// Initialize gateway
if config.OvnKubeNode.Mode == types.NodeModeDPUHost {
err = n.initGatewayDPUHost(nodeAddr)
if err != nil {
return err
}
} else {
if err := n.initGateway(subnets, nodeAnnotator, waiter, mgmtPortConfig, nodeAddr); err != nil {
return err
}
}
if err := nodeAnnotator.Run(); err != nil {
return fmt.Errorf("failed to set node %s annotations: %v", n.name, err)
}
// Wait for management port and gateway resources to be created by the master
klog.Infof("Waiting for gateway and management port readiness...")
start := time.Now()
if err := waiter.Wait(); err != nil {
return err
}
go n.gateway.Run(n.stopChan, wg)
klog.Infof("Gateway and management port readiness took %v", time.Since(start))
// Note(adrianc): DPU deployments are expected to support the new shared gateway changes, upgrade flow
// is not needed. Future upgrade flows will need to take DPUs into account.
if config.OvnKubeNode.Mode == types.NodeModeFull {
// Upgrade for Node. If we upgrade workers before masters, then we need to keep service routing via
// mgmt port until masters have been updated and modified OVN config. Run a goroutine to handle this case
// note this will change in the future to control-plane:
// https://github.com/kubernetes/kubernetes/pull/95382
masterNode, err := labels.NewRequirement("node-role.kubernetes.io/master", selection.Exists, nil)
if err != nil {
return err
}
labelSelector := labels.NewSelector()
labelSelector = labelSelector.Add(*masterNode)
informerFactory := informers.NewSharedInformerFactoryWithOptions(n.client, 0,
informers.WithTweakListOptions(func(options *metav1.ListOptions) {
options.LabelSelector = labelSelector.String()
}))
upgradeController := upgrade.NewController(n.Kube, informerFactory.Core().V1().Nodes())
initialTopoVersion := upgradeController.GetInitialTopoVersion()
bridgeName := n.gateway.GetGatewayBridgeIface()
needLegacySvcRoute := true
if (initialTopoVersion >= types.OvnHostToSvcOFTopoVersion && config.GatewayModeShared == config.Gateway.Mode) ||
(initialTopoVersion >= types.OvnRoutingViaHostTopoVersion) {
// Configure route for svc towards shared gw bridge
// Have to have the route to bridge for multi-NIC mode, where the default gateway may go to a non-OVS interface
if err := configureSvcRouteViaBridge(bridgeName); err != nil {
return err
}
needLegacySvcRoute = false
}
// Determine if we need to run upgrade checks
if initialTopoVersion != types.OvnCurrentTopologyVersion {
if needLegacySvcRoute {
klog.Info("System may be upgrading, falling back to to legacy K8S Service via mp0")
// add back legacy route for service via mp0
link, err := util.LinkSetUp(types.K8sMgmtIntfName)
if err != nil {
return fmt.Errorf("unable to get link for %s, error: %v", types.K8sMgmtIntfName, err)
}
var gwIP net.IP
for _, subnet := range config.Kubernetes.ServiceCIDRs {
if utilnet.IsIPv4CIDR(subnet) {
gwIP = mgmtPortConfig.ipv4.gwIP
} else {
gwIP = mgmtPortConfig.ipv6.gwIP
}
err := util.LinkRoutesAddOrUpdateMTU(link, gwIP, []*net.IPNet{subnet}, config.Default.RoutableMTU)
if err != nil {
return fmt.Errorf("unable to add legacy route for services via mp0, error: %v", err)
}
}
}
// need to run upgrade controller
informerStop := make(chan struct{})
informerFactory.Start(informerStop)
go func() {
if err := upgradeController.Run(n.stopChan, informerStop); err != nil {
klog.Fatalf("Error while running upgrade controller: %v", err)
}
// upgrade complete now see what needs upgrading
// migrate service route from ovn-k8s-mp0 to shared gw bridge
if (initialTopoVersion < types.OvnHostToSvcOFTopoVersion && config.GatewayModeShared == config.Gateway.Mode) ||
(initialTopoVersion < types.OvnRoutingViaHostTopoVersion) {
if err := upgradeServiceRoute(bridgeName); err != nil {
klog.Fatalf("Failed to upgrade service route for node, error: %v", err)
}
}
// ensure CNI support for port binding built into OVN, as masters have been upgraded
if initialTopoVersion < types.OvnPortBindingTopoVersion && cniServer != nil && !isOvnUpEnabled {
isOvnUpEnabled, err = getOVNIfUpCheckMode()
if err != nil {
klog.Errorf("%v", err)
}
if isOvnUpEnabled {
cniServer.EnableOVNPortUpSupport()
}
}
}()
}
}
if config.HybridOverlay.Enabled {
// Not supported with DPUs, enforced in config
// TODO(adrianc): Revisit above comment
nodeController, err := honode.NewNode(
n.Kube,
n.name,
n.watchFactory.NodeInformer(),
n.watchFactory.LocalPodInformer(),
informer.NewDefaultEventHandler,
)
if err != nil {
return err
}
wg.Add(1)
go func() {
defer wg.Done()
nodeController.Run(n.stopChan)
}()
}
if err := level.Set(strconv.Itoa(config.Logging.Level)); err != nil {
klog.Errorf("Reset of initial klog \"loglevel\" failed, err: %v", err)
}
// start management port health check
mgmtPort.CheckManagementPortHealth(mgmtPortConfig, n.stopChan)
if config.OvnKubeNode.Mode != types.NodeModeDPUHost {
// start health check to ensure there are no stale OVS internal ports
go wait.Until(func() {
checkForStaleOVSInterfaces(n.name, n.watchFactory.(*factory.WatchFactory))
}, time.Minute, n.stopChan)
util.SetARPTimeout()
n.WatchNamespaces()
// every minute cleanup stale conntrack entries if any
go wait.Until(func() {
n.checkAndDeleteStaleConntrackEntries()
}, time.Minute*1, n.stopChan)
n.WatchEndpoints()
}
if config.OvnKubeNode.Mode != types.NodeModeDPU {
// conditionally write cni config file
confFile := filepath.Join(config.CNI.ConfDir, config.CNIConfFileName)
_, err = os.Stat(confFile)
if os.IsNotExist(err) {
err = config.WriteCNIConfig()
if err != nil {
return err
}
}
}
if config.OvnKubeNode.Mode == types.NodeModeDPU {
n.watchPodsDPU(isOvnUpEnabled)
} else {
// start the cni server
err = cniServer.Start(cni.HandleCNIRequest)
}
return err
}
func (n *OvnNode) WatchEndpoints() {
n.watchFactory.AddEndpointsHandler(cache.ResourceEventHandlerFuncs{
UpdateFunc: func(old, new interface{}) {
epNew := new.(*kapi.Endpoints)
epOld := old.(*kapi.Endpoints)
newEpAddressMap := buildEndpointAddressMap(epNew.Subsets)
for item := range buildEndpointAddressMap(epOld.Subsets) {
if _, ok := newEpAddressMap[item]; !ok && item.protocol == kapi.ProtocolUDP { // flush conntrack only for UDP
err := util.DeleteConntrack(item.ip, item.port, item.protocol, netlink.ConntrackReplyAnyIP, nil)
if err != nil {
klog.Errorf("Failed to delete conntrack entry for %s: %v", item.ip, err)
}
}
}
},
DeleteFunc: func(obj interface{}) {
ep := obj.(*kapi.Endpoints)
for item := range buildEndpointAddressMap(ep.Subsets) {
if item.protocol == kapi.ProtocolUDP { // flush conntrack only for UDP
err := util.DeleteConntrack(item.ip, item.port, item.protocol, netlink.ConntrackReplyAnyIP, nil)
if err != nil {
klog.Errorf("Failed to delete conntrack entry for %s: %v", item.ip, err)
}
}
}
},
}, nil)
}
func exGatewayPodsAnnotationsChanged(oldNs, newNs *kapi.Namespace) bool {
// In reality we only care about exgw pod deletions, however since the list of IPs is not expected to change
// that often, let's check for *any* changes to these annotations compared to their previous state and trigger
// the logic for checking if we need to delete any conntrack entries
return (oldNs.Annotations[util.ExternalGatewayPodIPsAnnotation] != newNs.Annotations[util.ExternalGatewayPodIPsAnnotation]) ||
(oldNs.Annotations[util.RoutingExternalGWsAnnotation] != newNs.Annotations[util.RoutingExternalGWsAnnotation])
}
func (n *OvnNode) checkAndDeleteStaleConntrackEntries() {
namespaces, err := n.watchFactory.GetNamespaces()
if err != nil {
klog.Errorf("Unable to get pods from informer: %v", err)
}
for _, namespace := range namespaces {
_, foundRoutingExternalGWsAnnotation := namespace.Annotations[util.RoutingExternalGWsAnnotation]
_, foundExternalGatewayPodIPsAnnotation := namespace.Annotations[util.ExternalGatewayPodIPsAnnotation]
if foundRoutingExternalGWsAnnotation || foundExternalGatewayPodIPsAnnotation {
pods, err := n.watchFactory.GetPods(namespace.Name)
if err != nil {
klog.Warningf("Unable to get pods from informer for namespace %s: %v", namespace.Name, err)
}
if len(pods) > 0 || err != nil {
// we only need to proceed if there is at least one pod in this namespace on this node
// OR if we couldn't fetch the pods for some reason at this juncture
n.checkAndDeleteStaleConntrackEntriesForNamespace(namespace)
}
}
}
}
func (n *OvnNode) checkAndDeleteStaleConntrackEntriesForNamespace(newNs *kapi.Namespace) {
// loop through all the IPs on the annotations; ARP for their MACs and form an allowlist
gatewayIPs := strings.Split(newNs.Annotations[util.ExternalGatewayPodIPsAnnotation], ",")
gatewayIPs = append(gatewayIPs, strings.Split(newNs.Annotations[util.RoutingExternalGWsAnnotation], ",")...)
var wg sync.WaitGroup
wg.Add(len(gatewayIPs))
validMACs := sync.Map{}
for _, gwIP := range gatewayIPs {
go func(gwIP string) {
defer wg.Done()
if len(gwIP) > 0 {
if hwAddr, err := util.GetMACAddressFromARP(net.ParseIP(gwIP)); err != nil {
klog.Errorf("Failed to lookup hardware address for gatewayIP %s: %v", gwIP, err)
} else if len(hwAddr) > 0 {
// we need to reverse the mac before passing it to the conntrack filter since OVN saves the MAC in the following format
// +------------------------------------------------------------ +
// | 128 ... 112 ... 96 ... 80 ... 64 ... 48 ... 32 ... 16 ... 0|
// +------------------+-------+--------------------+-------------|
// | | UNUSED| MAC ADDRESS | UNUSED |
// +------------------+-------+--------------------+-------------+
for i, j := 0, len(hwAddr)-1; i < j; i, j = i+1, j-1 {
hwAddr[i], hwAddr[j] = hwAddr[j], hwAddr[i]
}
validMACs.Store(gwIP, []byte(hwAddr))
}
}
}(gwIP)
}
wg.Wait()
validNextHopMACs := [][]byte{}
validMACs.Range(func(key interface{}, value interface{}) bool {
validNextHopMACs = append(validNextHopMACs, value.([]byte))
return true
})
// Handle corner case where there are 0 IPs on the annotations OR none of the ARPs were successful; i.e allowMACList={empty}.
// This means we *need to* pass a label > 128 bits that will not match on any conntrack entry labels for these pods.
// That way any remaining entries with labels having MACs set will get purged.
if len(validNextHopMACs) == 0 {
validNextHopMACs = append(validNextHopMACs, []byte("does-not-contain-anything"))
}
pods, err := n.watchFactory.GetPods(newNs.Name)
if err != nil {
klog.Errorf("Unable to get pods from informer: %v", err)
}
for _, pod := range pods {
pod := pod
podIPs, err := util.GetAllPodIPs(pod)
if err != nil {
klog.Errorf("Unable to fetch IP for pod %s/%s: %v", pod.Namespace, pod.Name, err)
}
for _, podIP := range podIPs { // flush conntrack only for UDP
// for this pod, we check if the conntrack entry has a label that is not in the provided allowlist of MACs
// only caveat here is we assume egressGW served pods shouldn't have conntrack entries with other labels set
err := util.DeleteConntrack(podIP.String(), 0, kapi.ProtocolUDP, netlink.ConntrackOrigDstIP, validNextHopMACs)
if err != nil {
klog.Errorf("Failed to delete conntrack entry for pod %s: %v", podIP.String(), err)
}
}
}
}
func (n *OvnNode) WatchNamespaces() {
n.watchFactory.AddNamespaceHandler(cache.ResourceEventHandlerFuncs{
UpdateFunc: func(old, new interface{}) {
oldNs, newNs := old.(*kapi.Namespace), new.(*kapi.Namespace)
if exGatewayPodsAnnotationsChanged(oldNs, newNs) {
n.checkAndDeleteStaleConntrackEntriesForNamespace(newNs)
}
},
}, nil, n.watchFactory.GetHandlerPriority(""))
}
// validateVTEPInterfaceMTU checks if the MTU of the interface that has ovn-encap-ip is big
// enough to carry the `config.Default.MTU` and the Geneve header. If the MTU is not big
// enough, it will taint the node with the value of `types.OvnK8sSmallMTUTaintKey`
func (n *OvnNode) validateVTEPInterfaceMTU() error {
tooSmallMTUTaint := &kapi.Taint{Key: types.OvnK8sSmallMTUTaintKey, Effect: kapi.TaintEffectNoSchedule}
ovnEncapIP := net.ParseIP(config.Default.EncapIP)
if ovnEncapIP == nil {
return fmt.Errorf("the set OVN Encap IP is invalid: (%s)", config.Default.EncapIP)
}
interfaceName, mtu, err := util.GetIFNameAndMTUForAddress(ovnEncapIP)
if err != nil {
return fmt.Errorf("could not get MTU for the interface with address %s: %w", ovnEncapIP, err)
}
// calc required MTU
var requiredMTU int
if config.IPv4Mode && !config.IPv6Mode {
// we run in single-stack IPv4 only
requiredMTU = config.Default.MTU + types.GeneveHeaderLengthIPv4
} else {
// we run in single-stack IPv6 or dual-stack mode
requiredMTU = config.Default.MTU + types.GeneveHeaderLengthIPv6
}
// check if node needs to be tainted
if mtu < requiredMTU {
klog.V(2).Infof("MTU (%d) of network interface %s is not big enough to deal with Geneve "+
"header overhead (sum %d). Tainting node with %v...", mtu, interfaceName,
requiredMTU, tooSmallMTUTaint)
return retry.RetryOnConflict(retry.DefaultRetry, func() error {
return n.Kube.SetTaintOnNode(n.name, tooSmallMTUTaint)
})
}
klog.V(2).Infof("MTU (%d) of network interface %s is big enough to deal with Geneve header overhead (sum %d). "+
"Making sure node is not tainted with %v...", mtu, interfaceName, requiredMTU, tooSmallMTUTaint)
return retry.RetryOnConflict(retry.DefaultRetry, func() error {
return n.Kube.RemoveTaintFromNode(n.name, tooSmallMTUTaint)
})
}
type epAddressItem struct {
ip string
port int32
protocol kapi.Protocol
}
//buildEndpointAddressMap builds a map of all UDP and SCTP ports in the endpoint subset along with that port's IP address
func buildEndpointAddressMap(epSubsets []kapi.EndpointSubset) map[epAddressItem]struct{} {
epMap := make(map[epAddressItem]struct{})
for _, subset := range epSubsets {
for _, address := range subset.Addresses {
for _, port := range subset.Ports {
if port.Protocol == kapi.ProtocolUDP || port.Protocol == kapi.ProtocolSCTP {
epMap[epAddressItem{
ip: address.IP,
port: port.Port,
protocol: port.Protocol,
}] = struct{}{}
}
}
}
}
return epMap
}
func configureSvcRouteViaBridge(bridge string) error {
gwIPs, _, err := getGatewayNextHops()
if err != nil {
return fmt.Errorf("unable to get the gateway next hops, error: %v", err)
}
return configureSvcRouteViaInterface(bridge, gwIPs)
}
func upgradeServiceRoute(bridgeName string) error {
klog.Info("Updating K8S Service route")
// Flush old routes
link, err := util.LinkSetUp(types.K8sMgmtIntfName)
if err != nil {
return fmt.Errorf("unable to get link: %s, error: %v", types.K8sMgmtIntfName, err)
}
if err := util.LinkRoutesDel(link, config.Kubernetes.ServiceCIDRs); err != nil {
return fmt.Errorf("unable to delete routes on upgrade, error: %v", err)
}
// add route via OVS bridge
if err := configureSvcRouteViaBridge(bridgeName); err != nil {
return fmt.Errorf("unable to add svc route via OVS bridge interface, error: %v", err)
}
klog.Info("Successfully updated Kubernetes service route towards OVS")
// Clean up gw0 and local ovs bridge as best effort
if err := deleteLocalNodeAccessBridge(); err != nil {
klog.Warningf("Error while removing Local Node Access Bridge, error: %v", err)
}
// Clean up gw0 related IPTable rules as best effort.
for _, ip := range []string{types.V4NodeLocalNATSubnet, types.V6NodeLocalNATSubnet} {
_, IPNet, err := net.ParseCIDR(ip)
if err != nil {
klog.Errorf("Failed to LocalGatewayNATRules: %v", err)
}
rules := getLocalGatewayNATRules(types.LocalnetGatewayNextHopPort, IPNet)
if err := delIptRules(rules); err != nil {
klog.Errorf("Failed to LocalGatewayNATRules: %v", err)
}
}
return nil
}