forked from localhots/shezmu
-
Notifications
You must be signed in to change notification settings - Fork 0
/
shezmu.go
239 lines (203 loc) · 5.03 KB
/
shezmu.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
package shezmu
import (
"fmt"
"log"
"os"
"os/signal"
"runtime/debug"
"sync"
"syscall"
"time"
"github.com/Vivino/go-shezmu/stats"
)
// Shezmu is the master daemon.
type Shezmu struct {
DaemonStats stats.Publisher
Logger Logger
NumWorkers int
daemons []Daemon
queue chan *task
runtimeStats stats.Manager
wgWorkers sync.WaitGroup
wgSystem sync.WaitGroup
shutdownWorkers chan struct{}
shutdownSystem chan struct{}
}
// Actor is a function that could be executed by daemon workers.
type Actor func()
// Logger is the interface that implements minimal logging functions.
type Logger interface {
Printf(format string, v ...interface{})
Println(v ...interface{})
}
type task struct {
daemon Daemon
actor Actor
createdAt time.Time
system bool
name string
}
const (
// DefaultNumWorkers is the default number of workers that would process
// tasks.
DefaultNumWorkers = 100
)
// Summon creates a new instance of Shezmu.
func Summon() *Shezmu {
return &Shezmu{
DaemonStats: &stats.Void{},
Logger: log.New(os.Stdout, "", log.LstdFlags),
NumWorkers: DefaultNumWorkers,
queue: make(chan *task),
runtimeStats: stats.NewBasicStats(),
shutdownWorkers: make(chan struct{}),
shutdownSystem: make(chan struct{}),
}
}
// AddDaemon adds a new daemon.
func (s *Shezmu) AddDaemon(d Daemon) {
base := d.base()
base.self = d
base.queue = s.queue
base.logger = s.Logger
base.shutdown = s.shutdownSystem
s.daemons = append(s.daemons, d)
}
// ClearDaemons clears the list of added daemons. StopDaemons() function MUST be
// called before calling ClearDaemons().
func (s *Shezmu) ClearDaemons() {
s.daemons = []Daemon{}
}
// StartDaemons starts all registered daemons.
func (s *Shezmu) StartDaemons() {
s.Logger.Printf("Starting %d workers", s.NumWorkers)
for i := 0; i < s.NumWorkers; i++ {
go s.runWorker()
}
s.Logger.Println("Setting up daemons")
for _, d := range s.daemons {
s.setupDaemon(d)
}
}
// StopDaemons stops all running daemons.
func (s *Shezmu) StopDaemons() {
close(s.shutdownSystem)
for _, d := range s.daemons {
d.Shutdown()
}
s.wgSystem.Wait()
close(s.shutdownWorkers)
s.wgWorkers.Wait()
close(s.queue)
// Re-open closed channels to allow starting new deamons afterwards
s.shutdownSystem = make(chan struct{})
s.shutdownWorkers = make(chan struct{})
s.queue = make(chan *task)
fmt.Println(s.runtimeStats.Fetch(stats.Latency))
}
// HandleSignals handles signals.
func (s *Shezmu) HandleSignals() {
ch := make(chan os.Signal, 1)
signal.Notify(ch, syscall.SIGINT)
for {
switch sig := <-ch; sig {
case syscall.SIGINT:
s.StopDaemons()
return
default:
s.Logger.Printf("Signal ignored: %s", sig)
}
}
}
func (s *Shezmu) setupDaemon(d Daemon) {
defer func() {
if err := recover(); err != nil {
s.Logger.Printf("Failed to setup daemon %s due to process termination", d)
}
}()
s.queue <- &task{
daemon: d,
actor: d.Startup,
createdAt: time.Now(),
system: true,
name: "startup",
}
}
func (s *Shezmu) runWorker() {
s.wgWorkers.Add(1)
defer s.wgWorkers.Done()
defer func() {
if err := recover(); err != nil {
s.Logger.Printf("Worker crashed. Error: %v\n", err)
debug.PrintStack()
go s.runWorker() // Restarting worker
}
}()
for {
select {
case t := <-s.queue:
s.processTask(t)
case <-s.shutdownWorkers:
return
}
}
}
func (s *Shezmu) processTask(t *task) {
dur := time.Now().Sub(t.createdAt)
s.runtimeStats.Add(stats.Latency, dur)
if t.system {
s.processSystemTask(t)
} else {
s.processGeneralTask(t)
}
}
func (s *Shezmu) processSystemTask(t *task) {
// Abort starting a system task if shutdown was already called. Otherwise
// incrementing a wait group counter will cause a panic. This should be an
// extremely rare scenario when a system task crashes and tries to restart
// after a shutdown call.
select {
case <-s.shutdownSystem:
return
default:
}
s.wgSystem.Add(1)
defer s.wgSystem.Done()
defer func() {
if err := recover(); err != nil {
s.Logger.Printf("System task %s recovered from a panic\nError: %v\n", t, err)
debug.PrintStack()
t.createdAt = time.Now()
s.queue <- t // Restarting task
} else {
s.Logger.Printf("System task %s finished\n", t)
}
}()
s.Logger.Printf("Starting system task %s\n", t)
t.actor() // <--- ACTION STARTS HERE
}
func (s *Shezmu) processGeneralTask(t *task) {
defer func() {
if val := recover(); val != nil {
err := interfaceToError(val)
s.DaemonStats.Error(t.daemon.String())
t.daemon.base().handlePanic(err)
s.Logger.Printf("Daemon %s recovered from a panic\nError: %s\n", t.daemon, err.Error())
debug.PrintStack()
}
}()
defer func(start time.Time) {
dur := time.Now().Sub(start)
s.DaemonStats.Add(t.daemon.String(), dur)
}(time.Now())
t.actor() // <--- ACTION STARTS HERE
}
func (t *task) String() string {
return fmt.Sprintf("%s[%s]", t.daemon, t.name)
}
func interfaceToError(val interface{}) error {
if terr, ok := val.(error); ok {
return terr
}
return fmt.Errorf("%v", val)
}