This repository has been archived by the owner on Feb 27, 2022. It is now read-only.
/
scheduler.go
260 lines (211 loc) · 6.19 KB
/
scheduler.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
package balancer
import (
"time"
"sync"
"github.com/ReToCode/openshift-cross-cluster-loadbalancer/balancer/balancing"
"github.com/ReToCode/openshift-cross-cluster-loadbalancer/balancer/core"
"github.com/ReToCode/openshift-cross-cluster-loadbalancer/balancer/stats"
"github.com/sirupsen/logrus"
"os"
)
type StatsOperationAction int
const (
IncrementConnection StatsOperationAction = iota
DecrementConnection
IncrementRefused
)
type ElectRequest struct {
Context core.Context
Response chan core.RouterHost
Err chan error
}
type SafeClusters struct {
v map[string]*core.Cluster
mux sync.Mutex
}
// Scheduler handles:
// - The overall state (clusters, router hosts)
// - Health-Check-Results
// - Election of target router hosts
type Scheduler struct {
clusters SafeClusters
StatsHandler *stats.StatsHandler
healthCheckResults chan core.HealthCheckResult
elect chan ElectRequest
ResetStats chan bool
stop chan bool
}
func NewScheduler() *Scheduler {
return &Scheduler{
clusters: SafeClusters{v: map[string]*core.Cluster{}},
StatsHandler: stats.NewHandler(),
healthCheckResults: make(chan core.HealthCheckResult),
elect: make(chan ElectRequest),
ResetStats: make(chan bool),
stop: make(chan bool),
}
}
func (s *Scheduler) Start() {
s.StatsHandler.Start()
hostsPushTicket := time.NewTicker(2 * time.Second)
go func() {
for {
select {
case <-s.ResetStats:
s.resetStats()
case checkResult := <-s.healthCheckResults:
s.handleHealthCheckResults(checkResult)
case electReq := <-s.elect:
s.handleRouterHostElect(electReq)
case <-hostsPushTicket.C:
s.StatsHandler.RouterHosts <- s.routerHosts()
s.resetRefusedStats()
case <-s.stop:
logrus.Info("Stopping scheduler")
s.StatsHandler.Stop()
s.clusters.mux.Lock()
for _, c := range s.clusters.v {
c.Stop()
}
s.clusters.mux.Unlock()
return
}
}
}()
}
func (s *Scheduler) Stop() {
s.stop <- true
}
func (s *Scheduler) AddOrUpdateCluster(clusterKey string, data core.ClusterUpdate) {
s.clusters.mux.Lock()
if existing, exists := s.clusters.v[clusterKey]; exists {
s.updateCluster(existing, data)
} else {
s.addCluster(clusterKey, data)
}
s.clusters.mux.Unlock()
}
func (s *Scheduler) addCluster(clusterKey string, data core.ClusterUpdate) {
logrus.Infof("Added cluster: %v", clusterKey)
// Create the new cluster
s.clusters.v[clusterKey] = core.NewCluster(clusterKey, data.Routes)
// Add all router hosts to it
for _, rh := range data.RouterHosts {
s.addRouterHost(clusterKey, rh)
}
}
func (s *Scheduler) addRouterHost(clusterKey string, rh core.RouterHost) {
// For testing purposes, add an override for Router host IPs
ose1Debug := os.Getenv("OSE1_OVERRIDE")
ose2Debug := os.Getenv("OSE2_OVERRIDE")
if len(ose1Debug) > 0 && clusterKey == "ose1" {
rh.HostIP = ose1Debug
}
if len(ose2Debug) > 0 && clusterKey == "ose2" {
rh.HostIP = ose2Debug
}
newHost := core.NewRouterHost(rh.Name, rh.HostIP, rh.HTTPPort, rh.HTTPSPort, s.healthCheckResults, clusterKey)
logrus.Infof("New router host was added: %v to scheduler. %v", newHost.Name, newHost.HostIP)
s.clusters.v[clusterKey].RouterHosts[newHost.Name] = newHost
}
func (s *Scheduler) updateCluster(ecl *core.Cluster, data core.ClusterUpdate) {
// Update routes
ecl.Routes = data.Routes
// Add new routers
for _, rh := range data.RouterHosts {
if _, exists := ecl.RouterHosts[rh.Name]; !exists {
s.addRouterHost(ecl.Key, rh)
}
}
// Remove old routers
for _, erh := range ecl.RouterHosts {
if _, exists := data.RouterHosts[erh.Name]; !exists {
logrus.Infof("Router host %v no longer exists, deleting it from cluster", erh.Name)
delete(ecl.RouterHosts, erh.Name)
}
}
}
func (s *Scheduler) UpdateRouterStats(clusterKey string, routerHostKey string, action StatsOperationAction) {
s.clusters.mux.Lock()
defer s.clusters.mux.Unlock()
routerHost, ok := s.clusters.v[clusterKey].RouterHosts[routerHostKey]
if !ok {
logrus.Warn("Trying operation ", action, " on not tracked router host ip: ", routerHostKey)
return
}
switch action {
case IncrementRefused:
routerHost.LastState.RefusedConnections++
case IncrementConnection:
routerHost.LastState.ActiveConnections++
routerHost.LastState.TotalConnections++
case DecrementConnection:
routerHost.LastState.ActiveConnections--
default:
logrus.Warn("Don't know how to handle action ", action)
}
}
func (s *Scheduler) ElectRouterHostRequest(ctx core.Context) (*core.RouterHost, error) {
r := ElectRequest{ctx, make(chan core.RouterHost), make(chan error)}
// Send election request
s.elect <- r
select {
case err := <-r.Err:
return nil, err
case routerHost := <-r.Response:
return &routerHost, nil
}
}
func (s *Scheduler) routerHosts() []core.RouterHost {
s.clusters.mux.Lock()
defer s.clusters.mux.Unlock()
l := make([]core.RouterHost, 0)
for _, c := range s.clusters.v {
for _, rh := range c.RouterHosts {
l = append(l, *rh)
}
}
return l
}
func (s *Scheduler) resetStats() {
s.clusters.mux.Lock()
for _, cl := range s.clusters.v {
for _, rh := range cl.RouterHosts {
rh.LastState.TotalConnections = 0
}
}
s.clusters.mux.Unlock()
}
func (s *Scheduler) resetRefusedStats(){
s.clusters.mux.Lock()
for _, cl := range s.clusters.v {
for _, rh := range cl.RouterHosts {
rh.LastState.RefusedConnections = 0
}
}
s.clusters.mux.Unlock()
}
func (s *Scheduler) handleRouterHostElect(req ElectRequest) {
s.clusters.mux.Lock()
defer s.clusters.mux.Unlock()
rh, err := balancing.ElectRouterHost(req.Context, s.clusters.v)
if err != nil {
req.Err <- err
return
}
req.Response <- *rh
}
func (s *Scheduler) handleHealthCheckResults(res core.HealthCheckResult) {
s.clusters.mux.Lock()
defer s.clusters.mux.Unlock()
// Healthy > not healthy
if res.RouterHost.LastState.Healthy && !res.Healthy {
logrus.Warningf("Router host %v on %v degraded", res.RouterHost.Name, res.RouterHost.ClusterKey)
}
// Not healthy > healthy
if !res.RouterHost.LastState.Healthy && res.Healthy {
logrus.Infof("Router host %v on %v became healthy", res.RouterHost.Name, res.RouterHost.ClusterKey)
}
// Update state
res.RouterHost.LastState.Healthy = res.Healthy
}