-
Notifications
You must be signed in to change notification settings - Fork 273
/
azure_vmss_cache.go
341 lines (288 loc) · 10.7 KB
/
azure_vmss_cache.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
/*
Copyright 2020 The Kubernetes 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 provider
import (
"context"
"fmt"
"strings"
"sync"
"time"
"github.com/Azure/azure-sdk-for-go/services/compute/mgmt/2020-12-01/compute"
"github.com/Azure/go-autorest/autorest/to"
"k8s.io/apimachinery/pkg/util/sets"
"k8s.io/klog/v2"
azcache "sigs.k8s.io/cloud-provider-azure/pkg/cache"
"sigs.k8s.io/cloud-provider-azure/pkg/consts"
)
type vmssVirtualMachinesEntry struct {
resourceGroup string
vmssName string
instanceID string
virtualMachine *compute.VirtualMachineScaleSetVM
lastUpdate time.Time
}
type vmssEntry struct {
vmss *compute.VirtualMachineScaleSet
resourceGroup string
lastUpdate time.Time
}
type availabilitySetNodeEntry struct {
vmNames sets.String
nodeNames sets.String
vms []compute.VirtualMachine
}
func (ss *ScaleSet) newVMSSCache() (*azcache.TimedCache, error) {
getter := func(key string) (interface{}, error) {
localCache := &sync.Map{} // [vmasName]*vmssEntry
allResourceGroups, err := ss.GetResourceGroups()
if err != nil {
return nil, err
}
for _, resourceGroup := range allResourceGroups.List() {
allScaleSets, rerr := ss.VirtualMachineScaleSetsClient.List(context.Background(), resourceGroup)
if rerr != nil {
klog.Errorf("VirtualMachineScaleSetsClient.List failed: %v", rerr)
return nil, rerr.Error()
}
for i := range allScaleSets {
scaleSet := allScaleSets[i]
if scaleSet.Name == nil || *scaleSet.Name == "" {
klog.Warning("failed to get the name of VMSS")
continue
}
localCache.Store(*scaleSet.Name, &vmssEntry{
vmss: &scaleSet,
resourceGroup: resourceGroup,
lastUpdate: time.Now().UTC(),
})
}
}
return localCache, nil
}
if ss.Config.VmssCacheTTLInSeconds == 0 {
ss.Config.VmssCacheTTLInSeconds = consts.VMSSCacheTTLDefaultInSeconds
}
return azcache.NewTimedcache(time.Duration(ss.Config.VmssCacheTTLInSeconds)*time.Second, getter)
}
func extractVmssVMName(name string) (string, string, error) {
split := strings.SplitAfter(name, consts.VMSSNameSeparator)
if len(split) < 2 {
klog.V(3).Infof("Failed to extract vmssVMName %q", name)
return "", "", ErrorNotVmssInstance
}
ssName := strings.Join(split[0:len(split)-1], "")
// removing the trailing `vmssNameSeparator` since we used SplitAfter
ssName = ssName[:len(ssName)-1]
instanceID := split[len(split)-1]
return ssName, instanceID, nil
}
// getVMSSVMCache returns an *azcache.TimedCache and cache key for a VMSS (creating that cache if new).
func (ss *ScaleSet) getVMSSVMCache(resourceGroup, vmssName string) (string, *azcache.TimedCache, error) {
cacheKey := strings.ToLower(fmt.Sprintf("%s/%s", resourceGroup, vmssName))
if entry, ok := ss.vmssVMCache.Load(cacheKey); ok {
cache := entry.(*azcache.TimedCache)
return cacheKey, cache, nil
}
cache, err := ss.newVMSSVirtualMachinesCache(resourceGroup, vmssName, cacheKey)
if err != nil {
return "", nil, err
}
ss.vmssVMCache.Store(cacheKey, cache)
return cacheKey, cache, nil
}
// gcVMSSVMCache delete stale VMSS VMs caches from deleted VMSSes.
func (ss *ScaleSet) gcVMSSVMCache() error {
cached, err := ss.vmssCache.Get(consts.VMSSKey, azcache.CacheReadTypeUnsafe)
if err != nil {
return err
}
vmsses := cached.(*sync.Map)
removed := map[string]bool{}
ss.vmssVMCache.Range(func(key, value interface{}) bool {
cacheKey := key.(string)
vlistIdx := cacheKey[strings.LastIndex(cacheKey, "/")+1:]
if _, ok := vmsses.Load(vlistIdx); !ok {
removed[cacheKey] = true
}
return true
})
for key := range removed {
ss.vmssVMCache.Delete(key)
}
return nil
}
// newVMSSVirtualMachinesCache instantiates a new VMs cache for VMs belonging to the provided VMSS.
func (ss *ScaleSet) newVMSSVirtualMachinesCache(resourceGroupName, vmssName, cacheKey string) (*azcache.TimedCache, error) {
vmssVirtualMachinesCacheTTL := time.Duration(ss.Config.VmssVirtualMachinesCacheTTLInSeconds) * time.Second
getter := func(key string) (interface{}, error) {
localCache := &sync.Map{} // [nodeName]*vmssVirtualMachinesEntry
oldCache := make(map[string]vmssVirtualMachinesEntry)
if vmssCache, ok := ss.vmssVMCache.Load(cacheKey); ok {
// get old cache before refreshing the cache
cache := vmssCache.(*azcache.TimedCache)
entry, exists, err := cache.Store.GetByKey(cacheKey)
if err != nil {
return nil, err
}
if exists {
cached := entry.(*azcache.AzureCacheEntry).Data
if cached != nil {
virtualMachines := cached.(*sync.Map)
virtualMachines.Range(func(key, value interface{}) bool {
oldCache[key.(string)] = *value.(*vmssVirtualMachinesEntry)
return true
})
}
}
}
vms, err := ss.listScaleSetVMs(vmssName, resourceGroupName)
if err != nil {
return nil, err
}
for i := range vms {
vm := vms[i]
if vm.OsProfile == nil || vm.OsProfile.ComputerName == nil {
klog.Warningf("failed to get computerName for vmssVM (%q)", vmssName)
continue
}
computerName := strings.ToLower(*vm.OsProfile.ComputerName)
if vm.NetworkProfile == nil || vm.NetworkProfile.NetworkInterfaces == nil {
klog.Warningf("skip caching vmssVM %s since its network profile hasn't initialized yet (probably still under creating)", computerName)
continue
}
vmssVMCacheEntry := &vmssVirtualMachinesEntry{
resourceGroup: resourceGroupName,
vmssName: vmssName,
instanceID: to.String(vm.InstanceID),
virtualMachine: &vm,
lastUpdate: time.Now().UTC(),
}
// set cache entry to nil when the VM is under deleting.
if vm.VirtualMachineScaleSetVMProperties != nil &&
strings.EqualFold(to.String(vm.VirtualMachineScaleSetVMProperties.ProvisioningState), string(compute.ProvisioningStateDeleting)) {
klog.V(4).Infof("VMSS virtualMachine %q is under deleting, setting its cache to nil", computerName)
vmssVMCacheEntry.virtualMachine = nil
}
localCache.Store(computerName, vmssVMCacheEntry)
delete(oldCache, computerName)
}
// add old missing cache data with nil entries to prevent aggressive
// ARM calls during cache invalidation
for name, vmEntry := range oldCache {
// if the nil cache entry has existed for vmssVirtualMachinesCacheTTL in the cache
// then it should not be added back to the cache
if vmEntry.virtualMachine == nil && time.Since(vmEntry.lastUpdate) > vmssVirtualMachinesCacheTTL {
klog.V(5).Infof("ignoring expired entries from old cache for %s", name)
continue
}
lastUpdate := time.Now().UTC()
if vmEntry.virtualMachine == nil {
// if this is already a nil entry then keep the time the nil
// entry was first created, so we can cleanup unwanted entries
lastUpdate = vmEntry.lastUpdate
}
klog.V(5).Infof("adding old entries to new cache for %s", name)
localCache.Store(name, &vmssVirtualMachinesEntry{
resourceGroup: vmEntry.resourceGroup,
vmssName: vmEntry.vmssName,
instanceID: vmEntry.instanceID,
virtualMachine: nil,
lastUpdate: lastUpdate,
})
}
return localCache, nil
}
return azcache.NewTimedcache(vmssVirtualMachinesCacheTTL, getter)
}
func (ss *ScaleSet) deleteCacheForNode(nodeName string) error {
node, err := ss.getNodeIdentityByNodeName(nodeName, azcache.CacheReadTypeUnsafe)
if err != nil {
klog.Errorf("deleteCacheForNode(%s) failed with error: %v", nodeName, err)
return err
}
cacheKey, timedcache, err := ss.getVMSSVMCache(node.resourceGroup, node.vmssName)
if err != nil {
klog.Errorf("deleteCacheForNode(%s) failed with error: %v", nodeName, err)
return err
}
vmcache, err := timedcache.Get(cacheKey, azcache.CacheReadTypeUnsafe)
if err != nil {
klog.Errorf("deleteCacheForNode(%s) failed with error: %v", nodeName, err)
return err
}
virtualMachines := vmcache.(*sync.Map)
virtualMachines.Delete(nodeName)
if err := ss.gcVMSSVMCache(); err != nil {
klog.Errorf("deleteCacheForNode(%s) failed to gc stale vmss caches: %v", nodeName, err)
}
return nil
}
func (ss *ScaleSet) newAvailabilitySetNodesCache() (*azcache.TimedCache, error) {
getter := func(key string) (interface{}, error) {
vmNames := sets.NewString()
resourceGroups, err := ss.GetResourceGroups()
if err != nil {
return nil, err
}
vmList := make([]compute.VirtualMachine, 0)
for _, resourceGroup := range resourceGroups.List() {
vms, err := ss.Cloud.ListVirtualMachines(resourceGroup)
if err != nil {
return nil, fmt.Errorf("newAvailabilitySetNodesCache: failed to list vms in the resource group %s: %w", resourceGroup, err)
}
for _, vm := range vms {
if vm.Name != nil {
vmNames.Insert(to.String(vm.Name))
vmList = append(vmList, vm)
}
}
}
// store all the node names in the cluster when the cache data was created.
nodeNames, err := ss.GetNodeNames()
if err != nil {
return nil, err
}
localCache := availabilitySetNodeEntry{
vmNames: vmNames,
nodeNames: nodeNames,
vms: vmList,
}
return localCache, nil
}
if ss.Config.AvailabilitySetNodesCacheTTLInSeconds == 0 {
ss.Config.AvailabilitySetNodesCacheTTLInSeconds = consts.AvailabilitySetNodesCacheTTLDefaultInSeconds
}
return azcache.NewTimedcache(time.Duration(ss.Config.AvailabilitySetNodesCacheTTLInSeconds)*time.Second, getter)
}
func (ss *ScaleSet) isNodeManagedByAvailabilitySet(nodeName string, crt azcache.AzureCacheReadType) (bool, error) {
// Assume all nodes are managed by VMSS when DisableAvailabilitySetNodes is enabled.
if ss.DisableAvailabilitySetNodes {
klog.V(2).Infof("Assuming node %q is managed by VMSS since DisableAvailabilitySetNodes is set to true", nodeName)
return false, nil
}
cached, err := ss.availabilitySetNodesCache.Get(consts.AvailabilitySetNodesKey, crt)
if err != nil {
return false, err
}
cachedNodes := cached.(availabilitySetNodeEntry).nodeNames
// if the node is not in the cache, assume the node has joined after the last cache refresh and attempt to refresh the cache.
if !cachedNodes.Has(nodeName) {
klog.V(2).Infof("Node %s has joined the cluster since the last VM cache refresh, refreshing the cache", nodeName)
cached, err = ss.availabilitySetNodesCache.Get(consts.AvailabilitySetNodesKey, azcache.CacheReadTypeForceRefresh)
if err != nil {
return false, err
}
}
cachedVMs := cached.(availabilitySetNodeEntry).vmNames
return cachedVMs.Has(nodeName), nil
}