-
Notifications
You must be signed in to change notification settings - Fork 344
/
daemonset.go
468 lines (412 loc) · 15.2 KB
/
daemonset.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
/*
Copyright 2018 Heptio 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 daemonset
import (
"context"
"crypto/tls"
"fmt"
"time"
appsv1 "k8s.io/api/apps/v1"
v1 "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/client-go/kubernetes"
"github.com/pkg/errors"
"github.com/sirupsen/logrus"
"github.com/vmware-tanzu/sonobuoy/pkg/errlog"
"github.com/vmware-tanzu/sonobuoy/pkg/plugin"
"github.com/vmware-tanzu/sonobuoy/pkg/plugin/driver"
"github.com/vmware-tanzu/sonobuoy/pkg/plugin/driver/utils"
"github.com/vmware-tanzu/sonobuoy/pkg/plugin/manifest"
sonotime "github.com/vmware-tanzu/sonobuoy/pkg/time"
)
const (
// pollingInterval is the time between polls when monitoring the job status.
pollingInterval = 10 * time.Second
// defaultSleepSeconds is the time after the plugin finishes for which Sonobuoy will sleep.
// The sleep functions as a way to prevent the daemonset from restarting the container once the
// process completes. There is currently no way to have a "run-once daemonset". Defaults
// to sleeping forever.
defaultSleepSeconds = "-1"
)
// Plugin is a plugin driver that dispatches containers to each node,
// expecting each pod to report to the aggregator.
type Plugin struct {
driver.Base
}
// Ensure DaemonSetPlugin implements plugin.Interface
var _ plugin.Interface = &Plugin{}
// NewPlugin creates a new DaemonSet plugin from the given Plugin Definition
// and sonobuoy aggregator address.
func NewPlugin(dfn manifest.Manifest, namespace, sonobuoyImage, imagePullPolicy, imagePullSecrets string, customAnnotations map[string]string) *Plugin {
return &Plugin{
driver.Base{
Definition: dfn,
SessionID: utils.GetSessionID(),
Namespace: namespace,
SonobuoyImage: sonobuoyImage,
ImagePullPolicy: imagePullPolicy,
ImagePullSecrets: imagePullSecrets,
CustomAnnotations: customAnnotations,
CleanedUp: false,
},
}
}
// ExpectedResults returns the list of results expected for this daemonset.
func (p *Plugin) ExpectedResults(nodes []v1.Node) []plugin.ExpectedResult {
nodes = p.filterByNodeSelector(nodes)
ret := make([]plugin.ExpectedResult, 0, len(nodes))
for _, node := range nodes {
ret = append(ret, plugin.ExpectedResult{
NodeName: node.Name,
ResultType: p.GetName(),
Order: p.Definition.SonobuoyConfig.Order,
})
}
return ret
}
// filterByNodeSelector will filter the list of nodes to just the ones matching the nodeSelector of the plugin.
func (p *Plugin) filterByNodeSelector(nodes []v1.Node) []v1.Node {
ps := p.Base.Definition.PodSpec
if ps == nil {
return nodes
}
// Match either the nodeSelector field for the affinity.
var nodeSelector *v1.NodeSelector
if ps.Affinity != nil && ps.Affinity.NodeAffinity != nil {
nodeSelector = ps.Affinity.NodeAffinity.RequiredDuringSchedulingIgnoredDuringExecution
}
// Translate the nodeSelector field into a labels.Selector with requirements.
ls := labels.NewSelector()
nodeSelMap := ps.NodeSelector
reqs := labels.Requirements{}
for k, v := range nodeSelMap {
newReq, err := labels.NewRequirement(k, selection.Equals, []string{v})
if err != nil {
logrus.Errorf("Filtering by labelSelector had a bad requirement: %v", err)
return nodes
}
reqs = append(reqs, *newReq)
}
ls = ls.Add(reqs...)
if ls.Empty() && nodeSelector == nil {
logrus.Trace("Filtering by nodes had no requirements, returning all nodes")
return nodes
}
retNodes := []v1.Node{}
for _, node := range nodes {
logrus.Tracef("Filtering by labelSelector, checking node.GetLabels(): %v against %v", node, node.GetLabels())
// Split checks up to clarify logging/debugging.
if !ls.Empty() && ls.Matches(labels.Set(node.GetLabels())) {
logrus.Tracef("Matched labelSelctors")
retNodes = append(retNodes, node)
continue
} else {
logrus.Tracef("Did not match labelSelctors")
}
if nodeMatchesNodeSelector(&node, nodeSelector) {
logrus.Tracef("Matched affinity")
retNodes = append(retNodes, node)
continue
} else {
logrus.Tracef("Did not match labelSelctors")
}
}
return retNodes
}
func (p *Plugin) createDaemonSetDefinition(hostname string, cert *tls.Certificate, ownerPod *v1.Pod, progressPort, resultDir string) appsv1.DaemonSet {
ds := appsv1.DaemonSet{}
annotations := map[string]string{
"sonobuoy-driver": p.GetDriver(),
"sonobuoy-plugin": p.GetName(),
}
for k, v := range p.CustomAnnotations {
annotations[k] = v
}
labels := map[string]string{
"component": "sonobuoy",
"sonobuoy-component": "plugin",
"sonobuoy-plugin": p.GetName(),
"sonobuoy-run": p.SessionID,
"tier": "analysis",
}
ds.ObjectMeta = metav1.ObjectMeta{
Name: fmt.Sprintf("sonobuoy-%s-daemon-set-%s", p.GetName(), p.SessionID),
Namespace: p.Namespace,
Labels: labels,
Annotations: annotations,
OwnerReferences: []metav1.OwnerReference{
{
APIVersion: "v1",
Kind: "Pod",
Name: ownerPod.GetName(),
UID: ownerPod.GetUID(),
},
},
}
ds.Spec.Selector = &metav1.LabelSelector{
MatchLabels: map[string]string{
"sonobuoy-run": p.SessionID,
},
}
ds.Spec.Template.ObjectMeta.Labels = labels
ds.Spec.Template.ObjectMeta.Annotations = p.CustomAnnotations
var podSpec v1.PodSpec
if p.Definition.PodSpec != nil {
podSpec = p.Definition.PodSpec.PodSpec
} else {
podSpec = driver.DefaultPodSpec(p.GetDriver())
}
podSpec.Containers = append(podSpec.Containers,
p.Definition.Spec.Container,
p.CreateWorkerContainerDefintion(hostname, cert, []string{"/sonobuoy", "worker", "single-node", "--level=trace", "-v=6", "--logtostderr", "--sleep=" + defaultSleepSeconds}, []string{}, progressPort, resultDir, p.Definition.Spec.Container.Env),
)
if len(p.ImagePullSecrets) > 0 {
podSpec.ImagePullSecrets = append(podSpec.ImagePullSecrets, v1.LocalObjectReference{
Name: p.ImagePullSecrets,
})
}
podSpec.Volumes = append(podSpec.Volumes, v1.Volume{
Name: "results",
VolumeSource: v1.VolumeSource{
EmptyDir: &v1.EmptyDirVolumeSource{},
},
})
for _, v := range p.Definition.ExtraVolumes {
podSpec.Volumes = append(podSpec.Volumes, v.Volume)
}
ds.Spec.Template.Spec = podSpec
return ds
}
// Run dispatches worker pods according to the DaemonSet's configuration.
func (p *Plugin) Run(kubeclient kubernetes.Interface, hostname string, cert *tls.Certificate, ownerPod *v1.Pod, progressPort, resultDir string) error {
daemonSet := p.createDaemonSetDefinition(fmt.Sprintf("https://%s", hostname), cert, ownerPod, progressPort, resultDir)
secret, err := p.MakeTLSSecret(cert, ownerPod)
if err != nil {
return errors.Wrapf(err, "couldn't make secret for daemonset plugin %v", p.GetName())
}
if _, err := kubeclient.CoreV1().Secrets(p.Namespace).Create(context.TODO(), secret, metav1.CreateOptions{}); err != nil {
return errors.Wrapf(err, "couldn't create TLS secret for daemonset plugin %v", p.GetName())
}
if _, err := kubeclient.AppsV1().DaemonSets(p.Namespace).Create(context.TODO(), &daemonSet, metav1.CreateOptions{}); err != nil {
return errors.Wrapf(err, "could not create DaemonSet for daemonset plugin %v", p.GetName())
}
return nil
}
// Cleanup cleans up the k8s DaemonSet and ConfigMap created by this plugin instance.
func (p *Plugin) Cleanup(kubeclient kubernetes.Interface) {
p.CleanedUp = true
gracePeriod := int64(1)
deletionPolicy := metav1.DeletePropagationBackground
listOptions := p.listOptions()
deleteOptions := metav1.DeleteOptions{
GracePeriodSeconds: &gracePeriod,
PropagationPolicy: &deletionPolicy,
}
// Delete the DaemonSet created by this plugin
// TODO(EKF): Move to v1 in 1.11
err := kubeclient.AppsV1().DaemonSets(p.Namespace).DeleteCollection(
context.TODO(),
deleteOptions,
listOptions,
)
if err != nil {
errlog.LogError(errors.Wrapf(err, "could not delete DaemonSet-%v for daemonset plugin %v", p.GetSessionID(), p.GetName()))
}
}
func (p *Plugin) listOptions() metav1.ListOptions {
return metav1.ListOptions{
LabelSelector: "sonobuoy-run=" + p.GetSessionID(),
}
}
// findDaemonSet gets the daemonset that we created, using a kubernetes label search.
func (p *Plugin) findDaemonSet(kubeclient kubernetes.Interface) (*appsv1.DaemonSet, error) {
dsets, err := kubeclient.AppsV1().DaemonSets(p.Namespace).List(context.TODO(), p.listOptions())
if err != nil {
return nil, errors.WithStack(err)
}
if len(dsets.Items) != 1 {
return nil, errors.Errorf("expected plugin %v to create 1 daemonset, found %v", p.GetName(), len(dsets.Items))
}
return &dsets.Items[0], nil
}
// Monitor adheres to plugin.Interface by ensuring the DaemonSet is correctly
// configured and that each pod is running normally.
func (p *Plugin) Monitor(ctx context.Context, kubeclient kubernetes.Interface, availableNodes []v1.Node, resultsCh chan<- *plugin.Result) {
availableNodes = p.filterByNodeSelector(availableNodes)
podsReported := make(map[string]bool)
podsFound := make(map[string]bool, len(availableNodes))
for _, node := range availableNodes {
podsFound[node.Name] = false
podsReported[node.Name] = false
}
for {
// Sleep between each poll, which should give the DaemonSet
// enough time to create pods
select {
case <-ctx.Done():
switch {
// This routine is monitoring the entire daemonset and has no knowledge of which
// nodes have returned results to the aggregator. We can report the error for every node though
// since the aggregator will throw out duplicate results.
case ctx.Err() == context.DeadlineExceeded:
logrus.Errorf("Timeout waiting for plugin %v. Try checking the pod logs and other data in the results tarball for more information.", p.GetName())
errs := makeErrorResultsForNodes(
p.GetName(),
map[string]interface{}{"error": plugin.TimeoutErrMsg},
availableNodes,
)
for _, e := range errs {
resultsCh <- e
}
case ctx.Err() == context.Canceled:
// Do nothing, just stop.
case ctx.Err() != nil:
logrus.Errorf("Error seen while monitoring plugin %v: %v", p.GetName(), ctx.Err().Error())
errs := makeErrorResultsForNodes(
p.GetName(),
map[string]interface{}{"error": ctx.Err().Error()},
availableNodes,
)
for _, e := range errs {
resultsCh <- e
}
}
return
case <-sonotime.After(pollingInterval):
}
done, errResults := p.monitorOnce(kubeclient, availableNodes, podsFound, podsReported)
for _, v := range errResults {
resultsCh <- v
}
if done {
return
}
}
}
// monitorOnce handles the actual logic executed in the Monitor routine which also adds polling.
// It will return a boolean, indicating monitoring should stop, along with a result if one should
// be generated. The arguments, podsFound and podsReported, are used to persist some knowledge about
// the pods between calls.
func (p *Plugin) monitorOnce(kubeclient kubernetes.Interface, availableNodes []v1.Node, podsFound, podsReported map[string]bool) (done bool, retErrs []*plugin.Result) {
// If we've cleaned up after ourselves, stop monitoring
if p.CleanedUp {
return true, nil
}
// If we don't have a daemonset created, retry next time. We
// only send errors if we successfully see that an expected pod
// is having issues.
ds, err := p.findDaemonSet(kubeclient)
if err != nil {
errlog.LogError(errors.Wrapf(err, "could not find DaemonSet created by plugin %v, will retry", p.GetName()))
return false, nil
}
// Find all the pods configured by this daemonset
pods, err := kubeclient.CoreV1().Pods(p.Namespace).List(context.TODO(), p.listOptions())
if err != nil {
errlog.LogError(errors.Wrapf(err, "could not find pods created by plugin %v, will retry", p.GetName()))
// Likewise, if we can't query for pods, just retry next time.
return false, nil
}
// Cycle through each pod in this daemonset, reporting any failures.
for _, pod := range pods.Items {
nodeName := pod.Spec.NodeName
// We don't care about nodes we already saw
if podsReported[nodeName] {
continue
}
podsFound[nodeName] = true
// Check if it's failing and submit the error result
if isFailing, reason := utils.IsPodFailing(&pod); isFailing {
podsReported[nodeName] = true
retErrs = append(retErrs, utils.MakeErrorResult(p.GetName(), map[string]interface{}{
"error": reason,
"pod": pod,
}, nodeName))
}
}
// The main caller, Monitor, filters the list typically before passing it to us. However,
// the cost of the call is pretty small and this aids in testing to ensure that we do
// not error when the plugin isn't targeting that node.
availableNodes = p.filterByNodeSelector(availableNodes)
// DaemonSets are a bit strange, if node taints are preventing
// scheduling, pods won't even be created (unlike say Jobs,
// which will create the pod and leave it in an unscheduled
// state.) So take any nodes we didn't see pods on, and report
// issues scheduling them.
for _, node := range availableNodes {
if !podsFound[node.Name] && !podsReported[node.Name] {
podsReported[node.Name] = true
retErrs = append(retErrs, utils.MakeErrorResult(p.GetName(), map[string]interface{}{
"error": fmt.Sprintf(
"No pod was scheduled on node %v within %v. Check tolerations for plugin %v",
node.Name,
time.Since(ds.CreationTimestamp.Time),
p.GetName(),
),
}, node.Name))
}
}
return false, retErrs
}
func makeErrorResultsForNodes(resultType string, errdata map[string]interface{}, nodes []v1.Node) []*plugin.Result {
results := []*plugin.Result{}
for _, n := range nodes {
results = append(results, utils.MakeErrorResult(resultType, errdata, n.Name))
}
return results
}
// nodeMatchesNodeSelector checks if a node's labels satisfy a node selector. It is a simplification
// of upstream logic in `k8s.io/kubernetes` which isn't intended for consumption as a module.
// If the nodeSelector has multiple expressions/terms; this method returns the union of the nodes
// satisfying the individual terms.
func nodeMatchesNodeSelector(node *v1.Node, sel *v1.NodeSelector) bool {
if sel == nil || node == nil {
return false
}
for _, term := range sel.NodeSelectorTerms {
// We only support MatchExpressions at this time.
for _, exp := range term.MatchExpressions {
switch exp.Operator {
case v1.NodeSelectorOpExists:
if _, ok := node.Labels[exp.Key]; ok {
return true
}
case v1.NodeSelectorOpDoesNotExist:
if _, ok := node.Labels[exp.Key]; !ok {
return true
}
case v1.NodeSelectorOpIn:
if val, ok := node.Labels[exp.Key]; ok && stringInList(exp.Values, val) {
return true
}
case v1.NodeSelectorOpNotIn:
if val, ok := node.Labels[exp.Key]; !ok || !stringInList(exp.Values, val) {
return true
}
default:
continue
}
}
}
return false
}
func stringInList(list []string, s string) bool {
for _, v := range list {
if v == s {
return true
}
}
return false
}