-
-
Notifications
You must be signed in to change notification settings - Fork 740
/
heartbeat.go
202 lines (173 loc) · 5.2 KB
/
heartbeat.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
// Copyright 2020 Kentaro Hibino. All rights reserved.
// Use of this source code is governed by a MIT license
// that can be found in the LICENSE file.
package asynq
import (
"os"
"sync"
"time"
"github.com/google/uuid"
"github.com/hibiken/asynq/internal/base"
"github.com/hibiken/asynq/internal/log"
"github.com/hibiken/asynq/internal/timeutil"
)
// heartbeater is responsible for writing process info to redis periodically to
// indicate that the background worker process is up.
type heartbeater struct {
logger *log.Logger
broker base.Broker
clock timeutil.Clock
// channel to communicate back to the long running "heartbeater" goroutine.
done chan struct{}
// interval between heartbeats.
interval time.Duration
// following fields are initialized at construction time and are immutable.
host string
pid int
serverID string
concurrency int
queues map[string]int
strictPriority bool
// following fields are mutable and should be accessed only by the
// heartbeater goroutine. In other words, confine these variables
// to this goroutine only.
started time.Time
workers map[string]*workerInfo
// state is shared with other goroutine but is concurrency safe.
state *serverState
// channels to receive updates on active workers.
starting <-chan *workerInfo
finished <-chan *base.TaskMessage
}
type heartbeaterParams struct {
logger *log.Logger
broker base.Broker
interval time.Duration
concurrency int
queues map[string]int
strictPriority bool
state *serverState
starting <-chan *workerInfo
finished <-chan *base.TaskMessage
}
func newHeartbeater(params heartbeaterParams) *heartbeater {
host, err := os.Hostname()
if err != nil {
host = "unknown-host"
}
return &heartbeater{
logger: params.logger,
broker: params.broker,
clock: timeutil.NewRealClock(),
done: make(chan struct{}),
interval: params.interval,
host: host,
pid: os.Getpid(),
serverID: uuid.New().String(),
concurrency: params.concurrency,
queues: params.queues,
strictPriority: params.strictPriority,
state: params.state,
workers: make(map[string]*workerInfo),
starting: params.starting,
finished: params.finished,
}
}
func (h *heartbeater) shutdown() {
h.logger.Debug("Heartbeater shutting down...")
// Signal the heartbeater goroutine to stop.
h.done <- struct{}{}
}
// A workerInfo holds an active worker information.
type workerInfo struct {
// the task message the worker is processing.
msg *base.TaskMessage
// the time the worker has started processing the message.
started time.Time
// deadline the worker has to finish processing the task by.
deadline time.Time
// lease the worker holds for the task.
lease *base.Lease
}
func (h *heartbeater) start(wg *sync.WaitGroup) {
wg.Add(1)
go func() {
defer wg.Done()
h.started = h.clock.Now()
h.beat()
timer := time.NewTimer(h.interval)
for {
select {
case <-h.done:
if err := h.broker.ClearServerState(h.host, h.pid, h.serverID); err != nil {
h.logger.Errorf("Failed to clear server state: %v", err)
}
h.logger.Debug("Heartbeater done")
timer.Stop()
return
case <-timer.C:
h.beat()
timer.Reset(h.interval)
case w := <-h.starting:
h.workers[w.msg.ID] = w
case msg := <-h.finished:
delete(h.workers, msg.ID)
}
}
}()
}
// beat extends lease for workers and writes server/worker info to redis.
func (h *heartbeater) beat() {
h.state.mu.Lock()
srvStatus := h.state.value.String()
h.state.mu.Unlock()
info := base.ServerInfo{
Host: h.host,
PID: h.pid,
ServerID: h.serverID,
Concurrency: h.concurrency,
Queues: h.queues,
StrictPriority: h.strictPriority,
Status: srvStatus,
Started: h.started,
ActiveWorkerCount: len(h.workers),
}
var ws []*base.WorkerInfo
idsByQueue := make(map[string][]string)
for id, w := range h.workers {
ws = append(ws, &base.WorkerInfo{
Host: h.host,
PID: h.pid,
ServerID: h.serverID,
ID: id,
Type: w.msg.Type,
Queue: w.msg.Queue,
Payload: w.msg.Payload,
Started: w.started,
Deadline: w.deadline,
})
// Check lease before adding to the set to make sure not to extend the lease if the lease is already expired.
if w.lease.IsValid() {
idsByQueue[w.msg.Queue] = append(idsByQueue[w.msg.Queue], id)
} else {
w.lease.NotifyExpiration() // notify processor if the lease is expired
}
}
// Note: Set TTL to be long enough so that it won't expire before we write again
// and short enough to expire quickly once the process is shut down or killed.
if err := h.broker.WriteServerState(&info, ws, h.interval*2); err != nil {
h.logger.Errorf("Failed to write server state data: %v", err)
}
for qname, ids := range idsByQueue {
expirationTime, err := h.broker.ExtendLease(qname, ids...)
if err != nil {
h.logger.Errorf("Failed to extend lease for tasks %v: %v", ids, err)
continue
}
for _, id := range ids {
if l := h.workers[id].lease; !l.Reset(expirationTime) {
h.logger.Warnf("Lease reset failed for %s; lease deadline: %v", id, l.Deadline())
}
}
}
}