forked from hibiken/asynq
-
Notifications
You must be signed in to change notification settings - Fork 0
/
recoverer.go
126 lines (109 loc) · 3.37 KB
/
recoverer.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
// 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 (
"context"
"sync"
"time"
"github.com/hibiken/asynq/internal/base"
"github.com/hibiken/asynq/internal/errors"
"github.com/hibiken/asynq/internal/log"
)
type recoverer struct {
logger *log.Logger
broker base.Broker
retryDelayFunc RetryDelayFunc
isFailureFunc func(error) bool
// channel to communicate back to the long running "recoverer" goroutine.
done chan struct{}
// list of queues to check for deadline.
queues []string
// poll interval.
interval time.Duration
}
type recovererParams struct {
logger *log.Logger
broker base.Broker
queues []string
interval time.Duration
retryDelayFunc RetryDelayFunc
isFailureFunc func(error) bool
}
func newRecoverer(params recovererParams) *recoverer {
return &recoverer{
logger: params.logger,
broker: params.broker,
done: make(chan struct{}),
queues: params.queues,
interval: params.interval,
retryDelayFunc: params.retryDelayFunc,
isFailureFunc: params.isFailureFunc,
}
}
func (r *recoverer) shutdown() {
r.logger.Debug("Recoverer shutting down...")
// Signal the recoverer goroutine to stop polling.
r.done <- struct{}{}
}
func (r *recoverer) start(wg *sync.WaitGroup) {
wg.Add(1)
go func() {
defer wg.Done()
r.recover()
timer := time.NewTimer(r.interval)
for {
select {
case <-r.done:
r.logger.Debug("Recoverer done")
timer.Stop()
return
case <-timer.C:
r.recover()
timer.Reset(r.interval)
}
}
}()
}
// ErrLeaseExpired error indicates that the task failed because the worker working on the task
// could not extend its lease due to missing heartbeats. The worker may have crashed or got cutoff from the network.
var ErrLeaseExpired = errors.New("asynq: task lease expired")
func (r *recoverer) recover() {
r.recoverLeaseExpiredTasks()
r.recoverStaleAggregationSets()
}
func (r *recoverer) recoverLeaseExpiredTasks() {
// Get all tasks which have expired 30 seconds ago or earlier to accomodate certain amount of clock skew.
cutoff := time.Now().Add(-30 * time.Second)
msgs, err := r.broker.ListLeaseExpired(cutoff, r.queues...)
if err != nil {
r.logger.Warnf("recoverer: could not list lease expired tasks: %v", err)
return
}
for _, msg := range msgs {
if msg.Retried >= msg.Retry {
r.archive(msg, ErrLeaseExpired)
} else {
r.retry(msg, ErrLeaseExpired)
}
}
}
func (r *recoverer) recoverStaleAggregationSets() {
for _, qname := range r.queues {
if err := r.broker.ReclaimStaleAggregationSets(qname); err != nil {
r.logger.Warnf("recoverer: could not reclaim stale aggregation sets in queue %q: %v", qname, err)
}
}
}
func (r *recoverer) retry(msg *base.TaskMessage, err error) {
delay := r.retryDelayFunc(msg.Retried, err, NewTask(msg.Type, msg.Payload))
retryAt := time.Now().Add(delay)
if err := r.broker.Retry(context.Background(), msg, retryAt, err.Error(), r.isFailureFunc(err)); err != nil {
r.logger.Warnf("recoverer: could not retry lease expired task: %v", err)
}
}
func (r *recoverer) archive(msg *base.TaskMessage, err error) {
if err := r.broker.Archive(context.Background(), msg, err.Error()); err != nil {
r.logger.Warnf("recoverer: could not move task to archive: %v", err)
}
}