forked from Azure/aks-engine
/
upgrader.go
587 lines (496 loc) · 19.2 KB
/
upgrader.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
// Copyright (c) Microsoft Corporation. All rights reserved.
// Licensed under the MIT license.
package kubernetesupgrade
import (
"context"
"encoding/json"
"fmt"
"math/rand"
"time"
"github.com/Azure/aks-engine/pkg/api"
"github.com/Azure/aks-engine/pkg/armhelpers"
"github.com/Azure/aks-engine/pkg/armhelpers/utils"
"github.com/Azure/aks-engine/pkg/engine"
"github.com/Azure/aks-engine/pkg/engine/transform"
"github.com/Azure/aks-engine/pkg/i18n"
"github.com/Azure/aks-engine/pkg/operations"
"github.com/sirupsen/logrus"
v1 "k8s.io/api/core/v1"
)
// Upgrader holds information on upgrading an AKS cluster
type Upgrader struct {
Translator *i18n.Translator
logger *logrus.Entry
ClusterTopology
Client armhelpers.AKSEngineClient
kubeConfig string
stepTimeout *time.Duration
AKSEngineVersion string
}
type vmStatus int
const (
defaultTimeout = time.Minute * 20
vmStatusUpgraded vmStatus = iota
vmStatusNotUpgraded
vmStatusIgnored
)
type vmInfo struct {
name string
status vmStatus
}
// Init initializes an upgrader struct
func (ku *Upgrader) Init(translator *i18n.Translator, logger *logrus.Entry, clusterTopology ClusterTopology, client armhelpers.AKSEngineClient, kubeConfig string, stepTimeout *time.Duration, aksEngineVersion string) {
ku.Translator = translator
ku.logger = logger
ku.ClusterTopology = clusterTopology
ku.Client = client
ku.kubeConfig = kubeConfig
ku.stepTimeout = stepTimeout
ku.AKSEngineVersion = aksEngineVersion
}
// RunUpgrade runs the upgrade pipeline
func (ku *Upgrader) RunUpgrade() error {
ctx, cancel := context.WithTimeout(context.Background(), 90*time.Minute)
defer cancel()
if err := ku.upgradeMasterNodes(ctx); err != nil {
return err
}
if err := ku.upgradeAgentScaleSets(ctx); err != nil {
return err
}
return ku.upgradeAgentPools(ctx)
}
// Validate will run validation post upgrade
func (ku *Upgrader) Validate() error {
return nil
}
func (ku *Upgrader) upgradeMasterNodes(ctx context.Context) error {
if ku.ClusterTopology.DataModel.Properties.MasterProfile == nil {
return nil
}
ku.logger.Infof("Master nodes StorageProfile: %s", ku.ClusterTopology.DataModel.Properties.MasterProfile.StorageProfile)
// Upgrade Master VMs
templateMap, parametersMap, err := ku.generateUpgradeTemplate(ku.ClusterTopology.DataModel, ku.AKSEngineVersion)
if err != nil {
return ku.Translator.Errorf("error generating upgrade template: %s", err.Error())
}
ku.logger.Infof("Prepping master nodes for upgrade...")
transformer := &transform.Transformer{
Translator: ku.Translator,
}
if err := transformer.NormalizeResourcesForK8sMasterUpgrade(ku.logger, templateMap, ku.DataModel.Properties.MasterProfile.IsManagedDisks(), nil); err != nil {
ku.logger.Errorf(err.Error())
return err
}
upgradeMasterNode := UpgradeMasterNode{
Translator: ku.Translator,
logger: ku.logger,
}
upgradeMasterNode.TemplateMap = templateMap
upgradeMasterNode.ParametersMap = parametersMap
upgradeMasterNode.UpgradeContainerService = ku.ClusterTopology.DataModel
upgradeMasterNode.ResourceGroup = ku.ClusterTopology.ResourceGroup
upgradeMasterNode.SubscriptionID = ku.ClusterTopology.SubscriptionID
upgradeMasterNode.Client = ku.Client
upgradeMasterNode.kubeConfig = ku.kubeConfig
if ku.stepTimeout == nil {
upgradeMasterNode.timeout = defaultTimeout
} else {
upgradeMasterNode.timeout = *ku.stepTimeout
}
expectedMasterCount := ku.ClusterTopology.DataModel.Properties.MasterProfile.Count
mastersUpgradedCount := len(*ku.ClusterTopology.UpgradedMasterVMs)
mastersToUgradeCount := expectedMasterCount - mastersUpgradedCount
ku.logger.Infof("Total expected master count: %d", expectedMasterCount)
ku.logger.Infof("Master nodes that need to be upgraded: %d", mastersToUgradeCount)
ku.logger.Infof("Master nodes that have been upgraded: %d", mastersUpgradedCount)
ku.logger.Infof("Starting upgrade of master nodes...")
masterNodesInCluster := len(*ku.ClusterTopology.MasterVMs) + mastersUpgradedCount
ku.logger.Infof("masterNodesInCluster: %d", masterNodesInCluster)
if masterNodesInCluster > expectedMasterCount {
return ku.Translator.Errorf("Total count of master VMs: %d exceeded expected count: %d", masterNodesInCluster, expectedMasterCount)
}
upgradedMastersIndex := make(map[int]bool)
for _, vm := range *ku.ClusterTopology.UpgradedMasterVMs {
ku.logger.Infof("Master VM: %s is upgraded to expected orchestrator version", *vm.Name)
masterIndex, _ := utils.GetVMNameIndex(vm.StorageProfile.OsDisk.OsType, *vm.Name)
upgradedMastersIndex[masterIndex] = true
}
for _, vm := range *ku.ClusterTopology.MasterVMs {
ku.logger.Infof("Upgrading Master VM: %s", *vm.Name)
masterIndex, _ := utils.GetVMNameIndex(vm.StorageProfile.OsDisk.OsType, *vm.Name)
err := upgradeMasterNode.DeleteNode(vm.Name, false)
if err != nil {
ku.logger.Infof("Error deleting master VM: %s, err: %v", *vm.Name, err)
return err
}
err = upgradeMasterNode.CreateNode(ctx, "master", masterIndex)
if err != nil {
ku.logger.Infof("Error creating upgraded master VM: %s", *vm.Name)
return err
}
err = upgradeMasterNode.Validate(vm.Name)
if err != nil {
ku.logger.Infof("Error validating upgraded master VM: %s", *vm.Name)
return err
}
upgradedMastersIndex[masterIndex] = true
}
// This condition is possible if the previous upgrade operation failed during master
// VM upgrade when a master VM was deleted but creation of upgraded master did not run.
if masterNodesInCluster < expectedMasterCount {
ku.logger.Infof(
"Found missing master VMs in the cluster. Reconstructing names of missing master VMs for recreation during upgrade...")
}
mastersToCreate := expectedMasterCount - masterNodesInCluster
ku.logger.Infof("Expected master count: %d, Creating %d more master VMs", expectedMasterCount, mastersToCreate)
// NOTE: this is NOT completely idempotent because it assumes that
// the OS disk has been deleted
for i := 0; i < mastersToCreate; i++ {
masterIndexToCreate := 0
for upgradedMastersIndex[masterIndexToCreate] {
masterIndexToCreate++
}
ku.logger.Infof("Creating upgraded master VM with index: %d", masterIndexToCreate)
err = upgradeMasterNode.CreateNode(ctx, "master", masterIndexToCreate)
if err != nil {
ku.logger.Infof("Error creating upgraded master VM with index: %d", masterIndexToCreate)
return err
}
tempVMName := ""
err = upgradeMasterNode.Validate(&tempVMName)
if err != nil {
ku.logger.Infof("Error validating upgraded master VM with index: %d", masterIndexToCreate)
return err
}
upgradedMastersIndex[masterIndexToCreate] = true
}
return nil
}
func (ku *Upgrader) upgradeAgentPools(ctx context.Context) error {
for _, agentPool := range ku.ClusterTopology.AgentPools {
// Upgrade Agent VMs
templateMap, parametersMap, err := ku.generateUpgradeTemplate(ku.ClusterTopology.DataModel, ku.AKSEngineVersion)
if err != nil {
ku.logger.Errorf("Error generating upgrade template: %v", err)
return ku.Translator.Errorf("Error generating upgrade template: %s", err.Error())
}
ku.logger.Infof("Prepping agent pool '%s' for upgrade...", *agentPool.Name)
preservePools := map[string]bool{*agentPool.Name: true}
transformer := &transform.Transformer{
Translator: ku.Translator,
}
var isMasterManagedDisk bool
if ku.DataModel.Properties.MasterProfile != nil {
isMasterManagedDisk = ku.DataModel.Properties.MasterProfile.IsManagedDisks()
}
if err := transformer.NormalizeResourcesForK8sAgentUpgrade(ku.logger, templateMap, isMasterManagedDisk, preservePools); err != nil {
ku.logger.Errorf(err.Error())
return ku.Translator.Errorf("Error generating upgrade template: %s", err.Error())
}
var agentCount, agentPoolIndex int
for indx, app := range ku.ClusterTopology.DataModel.Properties.AgentPoolProfiles {
if app.Name == *agentPool.Name {
agentCount = app.Count
agentPoolIndex = indx
break
}
}
if agentCount == 0 {
ku.logger.Infof("Agent pool '%s' is empty", *agentPool.Name)
return nil
}
upgradeAgentNode := UpgradeAgentNode{
Translator: ku.Translator,
logger: ku.logger,
}
upgradeAgentNode.TemplateMap = templateMap
upgradeAgentNode.ParametersMap = parametersMap
upgradeAgentNode.UpgradeContainerService = ku.ClusterTopology.DataModel
upgradeAgentNode.SubscriptionID = ku.ClusterTopology.SubscriptionID
upgradeAgentNode.ResourceGroup = ku.ClusterTopology.ResourceGroup
upgradeAgentNode.Client = ku.Client
upgradeAgentNode.kubeConfig = ku.kubeConfig
if ku.stepTimeout == nil {
upgradeAgentNode.timeout = defaultTimeout
} else {
upgradeAgentNode.timeout = *ku.stepTimeout
}
agentVMs := make(map[int]*vmInfo)
// Go over upgraded VMs and verify provisioning state
// per https://docs.microsoft.com/en-us/rest/api/compute/virtualmachines/virtualmachines-state :
// - Creating: Indicates the virtual Machine is being created.
// - Updating: Indicates that there is an update operation in progress on the Virtual Machine.
// - Succeeded: Indicates that the operation executed on the virtual machine succeeded.
// - Deleting: Indicates that the virtual machine is being deleted.
// - Failed: Indicates that the update operation on the Virtual Machine failed.
// Delete VMs in 'bad' state. Such VMs will be re-created later in this function.
upgradedCount := 0
for _, vm := range *agentPool.UpgradedAgentVMs {
ku.logger.Infof("Agent VM: %s, pool name: %s on expected orchestrator version", *vm.Name, *agentPool.Name)
var vmProvisioningState string
if vm.VirtualMachineProperties != nil && vm.VirtualMachineProperties.ProvisioningState != nil {
vmProvisioningState = *vm.VirtualMachineProperties.ProvisioningState
}
agentIndex, _ := utils.GetVMNameIndex(vm.StorageProfile.OsDisk.OsType, *vm.Name)
switch vmProvisioningState {
case "Creating", "Updating", "Succeeded":
agentVMs[agentIndex] = &vmInfo{*vm.Name, vmStatusUpgraded}
upgradedCount++
case "Failed":
ku.logger.Infof("Deleting agent VM %s in provisioning state %s", *vm.Name, vmProvisioningState)
err := upgradeAgentNode.DeleteNode(vm.Name, false)
if err != nil {
ku.logger.Errorf("Error deleting agent VM %s: %v", *vm.Name, err)
return err
}
case "Deleting":
fallthrough
default:
ku.logger.Infof("Ignoring agent VM %s in provisioning state %s", *vm.Name, vmProvisioningState)
agentVMs[agentIndex] = &vmInfo{*vm.Name, vmStatusIgnored}
}
}
for _, vm := range *agentPool.AgentVMs {
agentIndex, _ := utils.GetVMNameIndex(vm.StorageProfile.OsDisk.OsType, *vm.Name)
agentVMs[agentIndex] = &vmInfo{*vm.Name, vmStatusNotUpgraded}
}
toBeUpgradedCount := len(*agentPool.AgentVMs)
ku.logger.Infof("Starting upgrade of %d agent nodes (out of %d) in pool identifier: %s, name: %s...",
toBeUpgradedCount, agentCount, *agentPool.Identifier, *agentPool.Name)
// Create missing nodes to match agentCount. This could be due to previous upgrade failure
// If there are nodes that need to be upgraded, create one extra node, which will be used to take on the load from upgrading nodes.
if toBeUpgradedCount > 0 {
agentCount++
}
for upgradedCount+toBeUpgradedCount < agentCount {
agentIndex := getAvailableIndex(agentVMs)
vmName, err := utils.GetK8sVMName(ku.DataModel.Properties, agentPoolIndex, agentIndex)
if err != nil {
ku.logger.Errorf("Error reconstructing agent VM name with index %d: %v", agentIndex, err)
return err
}
ku.logger.Infof("Creating new agent node %s (index %d)", vmName, agentIndex)
err = upgradeAgentNode.CreateNode(ctx, *agentPool.Name, agentIndex)
if err != nil {
ku.logger.Errorf("Error creating agent node %s (index %d): %v", vmName, agentIndex, err)
return err
}
err = upgradeAgentNode.Validate(&vmName)
if err != nil {
ku.logger.Infof("Error validating agent node %s (index %d): %v", vmName, agentIndex, err)
return err
}
agentVMs[agentIndex] = &vmInfo{vmName, vmStatusUpgraded}
upgradedCount++
}
if toBeUpgradedCount == 0 {
ku.logger.Infof("No nodes to upgrade")
return nil
}
// Upgrade nodes in agent pool
upgradedCount = 0
for agentIndex, vm := range agentVMs {
if vm.status != vmStatusNotUpgraded {
continue
}
ku.logger.Infof("Upgrading Agent VM: %s, pool name: %s", vm.name, *agentPool.Name)
err := upgradeAgentNode.DeleteNode(&vm.name, true)
if err != nil {
ku.logger.Errorf("Error deleting agent VM %s: %v", vm.name, err)
return err
}
vmName, err := utils.GetK8sVMName(ku.DataModel.Properties, agentPoolIndex, agentIndex)
if err != nil {
ku.logger.Errorf("Error fetching new VM name: %v", err)
return err
}
// do not create last node in favor of already created extra node.
if upgradedCount == toBeUpgradedCount-1 {
ku.logger.Infof("Skipping creation of VM %s (index %d)", vmName, agentIndex)
delete(agentVMs, agentIndex)
} else {
err = upgradeAgentNode.CreateNode(ctx, *agentPool.Name, agentIndex)
if err != nil {
ku.logger.Errorf("Error creating upgraded agent VM %s: %v", vmName, err)
return err
}
err = upgradeAgentNode.Validate(&vmName)
if err != nil {
ku.logger.Errorf("Error validating upgraded agent VM %s: %v", vmName, err)
return err
}
vm.status = vmStatusUpgraded
}
upgradedCount++
}
}
return nil
}
func (ku *Upgrader) upgradeAgentScaleSets(ctx context.Context) error {
if len(ku.ClusterTopology.AgentPoolScaleSetsToUpgrade) > 0 {
// need to apply the ARM template with target Kubernetes version to the VMSS first in order that the new VMSS instances
// created can get the expected Kubernetes version. Otherwise the new instances created still have old Kubernetes version
// if the topology doesn't have master nodes (so there are no ARM deployments in previous upgradeMasterNodes step)
templateMap, parametersMap, err := ku.generateUpgradeTemplate(ku.ClusterTopology.DataModel, ku.AKSEngineVersion)
if err != nil {
ku.logger.Errorf("error generating upgrade template in upgradeAgentScaleSets: %v", err)
return err
}
transformer := &transform.Transformer{
Translator: ku.Translator,
}
if err := transformer.NormalizeForVMSSScaling(ku.logger, templateMap); err != nil {
ku.logger.Errorf("unable to update template, error: %v.", err)
return err
}
random := rand.New(rand.NewSource(time.Now().UnixNano()))
deploymentSuffix := random.Int31()
deploymentName := fmt.Sprintf("agentscaleset-%s-%d", time.Now().Format("06-01-02T15.04.05"), deploymentSuffix)
ku.logger.Infof("Deploying the agent scale sets ARM template...")
_, err = ku.Client.DeployTemplate(
ctx,
ku.ClusterTopology.ResourceGroup,
deploymentName,
templateMap,
parametersMap)
if err != nil {
ku.logger.Errorf("error applying upgrade template in upgradeAgentScaleSets: %v", err)
return err
}
}
for _, vmssToUpgrade := range ku.ClusterTopology.AgentPoolScaleSetsToUpgrade {
ku.logger.Infof("Upgrading VMSS %s", vmssToUpgrade.Name)
if len(vmssToUpgrade.VMsToUpgrade) == 0 {
ku.logger.Infof("No VMs to upgrade for VMSS %s, skipping", vmssToUpgrade.Name)
continue
}
newCapacity := *vmssToUpgrade.Sku.Capacity + 1
ku.logger.Infof(
"VMSS %s current capacity is %d and new capacity will be %d while each node is swapped",
vmssToUpgrade.Name,
*vmssToUpgrade.Sku.Capacity,
newCapacity,
)
*vmssToUpgrade.Sku.Capacity = newCapacity
for _, vmToUpgrade := range vmssToUpgrade.VMsToUpgrade {
if err := ku.Client.SetVirtualMachineScaleSetCapacity(
ctx,
ku.ClusterTopology.ResourceGroup,
vmssToUpgrade.Name,
vmssToUpgrade.Sku,
vmssToUpgrade.Location,
); err != nil {
ku.logger.Errorf("Failure to set capacity for VMSS %s", vmssToUpgrade.Name)
return err
}
ku.logger.Infof("Successfully set capacity for VMSS %s", vmssToUpgrade.Name)
// Before we can delete the node we should safely and responsibly drain it
var kubeAPIServerURL string
getClientTimeout := 10 * time.Second
if ku.DataModel.Properties.HostedMasterProfile != nil {
kubeAPIServerURL = ku.DataModel.Properties.HostedMasterProfile.FQDN
} else {
kubeAPIServerURL = ku.DataModel.Properties.MasterProfile.FQDN
}
client, err := ku.Client.GetKubernetesClient(
kubeAPIServerURL,
ku.kubeConfig,
interval,
getClientTimeout,
)
if err != nil {
ku.logger.Errorf("Error getting Kubernetes client: %v", err)
return err
}
ku.logger.Infof("Draining node %s", vmToUpgrade.Name)
err = operations.SafelyDrainNodeWithClient(
client,
ku.logger,
vmToUpgrade.Name,
time.Minute,
)
if err != nil {
ku.logger.Errorf("Error draining VM in VMSS: %v", err)
return err
}
ku.logger.Infof(
"Deleting VM %s in VMSS %s",
vmToUpgrade.Name,
vmssToUpgrade.Name,
)
// At this point we have our buffer node that will replace the node to delete
// so we can just remove this current node then
if err := ku.Client.DeleteVirtualMachineScaleSetVM(
ctx,
ku.ClusterTopology.ResourceGroup,
vmssToUpgrade.Name,
vmToUpgrade.InstanceID,
); err != nil {
ku.logger.Errorf(
"Failed to delete VM %s in VMSS %s",
vmToUpgrade.Name,
vmssToUpgrade.Name)
return err
}
ku.logger.Infof(
"Successfully deleted VM %s in VMSS %s",
vmToUpgrade.Name,
vmssToUpgrade.Name)
}
ku.logger.Infof("Completed upgrading VMSS %s", vmssToUpgrade.Name)
}
ku.logger.Infoln("Completed upgrading all VMSS")
return nil
}
func (ku *Upgrader) generateUpgradeTemplate(upgradeContainerService *api.ContainerService, aksEngineVersion string) (map[string]interface{}, map[string]interface{}, error) {
var err error
ctx := engine.Context{
Translator: ku.Translator,
}
templateGenerator, err := engine.InitializeTemplateGenerator(ctx)
if err != nil {
return nil, nil, ku.Translator.Errorf("failed to initialize template generator: %s", err.Error())
}
_, err = upgradeContainerService.SetPropertiesDefaults(true, false)
if err != nil {
return nil, nil, ku.Translator.Errorf("error in SetPropertiesDefaults: %s", err.Error())
}
var templateJSON string
var parametersJSON string
if templateJSON, parametersJSON, err = templateGenerator.GenerateTemplate(upgradeContainerService, engine.DefaultGeneratorCode, aksEngineVersion); err != nil {
return nil, nil, ku.Translator.Errorf("error generating upgrade template: %s", err.Error())
}
var template interface{}
var parameters interface{}
json.Unmarshal([]byte(templateJSON), &template)
json.Unmarshal([]byte(parametersJSON), ¶meters)
templateMap := template.(map[string]interface{})
parametersMap := parameters.(map[string]interface{})
return templateMap, parametersMap, nil
}
// return unused index within the range of agent indices, or subsequent index
func getAvailableIndex(vms map[int]*vmInfo) int {
maxIndex := 0
for indx := range vms {
if indx > maxIndex {
maxIndex = indx
}
}
for indx := 0; indx < maxIndex; indx++ {
if _, found := vms[indx]; !found {
return indx
}
}
return maxIndex + 1
}
// isNodeReady returns true if a node is ready; false otherwise.
// Copied from: https://github.com/kubernetes/kubernetes/blob/886e04f1fffbb04faf8a9f9ee141143b2684ae68/pkg/api/v1/node/util.go#L40
func isNodeReady(node *v1.Node) bool {
for _, c := range node.Status.Conditions {
if c.Type == v1.NodeReady {
return c.Status == v1.ConditionTrue
}
}
return false
}