generated from kubernetes/kubernetes-template-project
-
Notifications
You must be signed in to change notification settings - Fork 44
/
Copy pathprovider.go
175 lines (155 loc) · 4.04 KB
/
provider.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
package backend
import (
"context"
"fmt"
"sync"
"time"
"go.uber.org/multierr"
klog "k8s.io/klog/v2"
)
const (
fetchMetricsTimeout = 5 * time.Second
)
func NewProvider(pmc PodMetricsClient, datastore *K8sDatastore) *Provider {
p := &Provider{
podMetrics: sync.Map{},
pmc: pmc,
datastore: datastore,
}
return p
}
// Provider provides backend pods and information such as metrics.
type Provider struct {
// key: Pod, value: *PodMetrics
podMetrics sync.Map
pmc PodMetricsClient
datastore *K8sDatastore
}
type PodMetricsClient interface {
FetchMetrics(ctx context.Context, pod Pod, existing *PodMetrics) (*PodMetrics, error)
}
func (p *Provider) AllPodMetrics() []*PodMetrics {
res := []*PodMetrics{}
fn := func(k, v any) bool {
res = append(res, v.(*PodMetrics))
return true
}
p.podMetrics.Range(fn)
return res
}
func (p *Provider) UpdatePodMetrics(pod Pod, pm *PodMetrics) {
p.podMetrics.Store(pod, pm)
}
func (p *Provider) GetPodMetrics(pod Pod) (*PodMetrics, bool) {
val, ok := p.podMetrics.Load(pod)
if ok {
return val.(*PodMetrics), true
}
return nil, false
}
func (p *Provider) Init(refreshPodsInterval, refreshMetricsInterval time.Duration) error {
p.refreshPodsOnce()
if err := p.refreshMetricsOnce(); err != nil {
klog.Errorf("Failed to init metrics: %v", err)
}
klog.Infof("Initialized pods and metrics: %+v", p.AllPodMetrics())
// periodically refresh pods
go func() {
for {
time.Sleep(refreshPodsInterval)
p.refreshPodsOnce()
}
}()
// periodically refresh metrics
go func() {
for {
time.Sleep(refreshMetricsInterval)
if err := p.refreshMetricsOnce(); err != nil {
klog.V(4).Infof("Failed to refresh metrics: %v", err)
}
}
}()
// Periodically print out the pods and metrics for DEBUGGING.
if klog.V(2).Enabled() {
go func() {
for {
time.Sleep(5 * time.Second)
klog.Infof("===DEBUG: Current Pods and metrics: %+v", p.AllPodMetrics())
}
}()
}
return nil
}
// refreshPodsOnce lists pods and updates keys in the podMetrics map.
// Note this function doesn't update the PodMetrics value, it's done separately.
func (p *Provider) refreshPodsOnce() {
// merge new pods with cached ones.
// add new pod to the map
addNewPods := func(k, v any) bool {
pod := k.(Pod)
if _, ok := p.podMetrics.Load(pod); !ok {
new := &PodMetrics{
Pod: pod,
Metrics: Metrics{
ActiveModels: make(map[string]int),
},
}
p.podMetrics.Store(pod, new)
}
return true
}
// remove pods that don't exist any more.
mergeFn := func(k, v any) bool {
pod := k.(Pod)
if _, ok := p.datastore.pods.Load(pod); !ok {
p.podMetrics.Delete(pod)
}
return true
}
p.podMetrics.Range(mergeFn)
p.datastore.pods.Range(addNewPods)
}
func (p *Provider) refreshMetricsOnce() error {
ctx, cancel := context.WithTimeout(context.Background(), fetchMetricsTimeout)
defer cancel()
start := time.Now()
defer func() {
d := time.Since(start)
// TODO: add a metric instead of logging
klog.V(4).Infof("Refreshed metrics in %v", d)
}()
var wg sync.WaitGroup
errCh := make(chan error)
processOnePod := func(key, value any) bool {
klog.V(4).Infof("Processing pod %v and metric %v", key, value)
pod := key.(Pod)
existing := value.(*PodMetrics)
wg.Add(1)
go func() {
defer wg.Done()
updated, err := p.pmc.FetchMetrics(ctx, pod, existing)
if err != nil {
errCh <- fmt.Errorf("failed to parse metrics from %s: %v", pod, err)
return
}
p.UpdatePodMetrics(pod, updated)
klog.V(4).Infof("Updated metrics for pod %s: %v", pod, updated.Metrics)
}()
return true
}
p.podMetrics.Range(processOnePod)
// Wait for metric collection for all pods to complete and close the error channel in a
// goroutine so this is unblocking, allowing the code to proceed to the error collection code
// below.
// Note we couldn't use a buffered error channel with a size because the size of the podMetrics
// sync.Map is unknown beforehand.
go func() {
wg.Wait()
close(errCh)
}()
var errs error
for err := range errCh {
errs = multierr.Append(errs, err)
}
return errs
}