-
Notifications
You must be signed in to change notification settings - Fork 14
/
delayconsumer.go
238 lines (209 loc) · 8.11 KB
/
delayconsumer.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
package retry
import (
"context"
"github.com/razorpay/metro/internal/brokerstore"
"github.com/razorpay/metro/internal/subscription"
topic "github.com/razorpay/metro/internal/topic"
"github.com/razorpay/metro/pkg/cache"
"github.com/razorpay/metro/pkg/logger"
"github.com/razorpay/metro/pkg/messagebroker"
)
// DelayConsumer ...
type DelayConsumer struct {
subscriberID string
ctx context.Context
cancelFunc func()
doneCh chan struct{}
subs *subscription.Model
topic string
isPaused bool
consumer messagebroker.Consumer
bs brokerstore.IBrokerStore
handler MessageHandler
ch cache.ICache
// a paused consumer will not return new messages, so this cachedMsg will be used for lookups
// till the needed time elapses
cachedMsgs []messagebroker.ReceivedMessage
errChan chan error
}
// NewDelayConsumer inits a new delay-consumer with the pre-defined message handler
func NewDelayConsumer(ctx context.Context, subscriberID, topicName string, subs *subscription.Model, bs brokerstore.IBrokerStore, handler MessageHandler, ch cache.ICache, errChan chan error) (*DelayConsumer, error) {
delayCtx, cancel := context.WithCancel(ctx)
// only delay-consumer will consume from a subscription specific delay-topic, so can use the same groupID and groupInstanceID
consumerOps := messagebroker.ConsumerClientOptions{
Topics: []string{topicName},
GroupID: subs.GetDelayConsumerGroupID(topicName),
GroupInstanceID: subs.GetDelayConsumerGroupInstanceID(subscriberID, topicName),
}
consumer, err := bs.GetConsumer(ctx, consumerOps)
if err != nil {
logger.Ctx(ctx).Errorw("delay-consumer: failed to create consumer", "error", err.Error())
return nil, err
}
// on init, make sure to call resume. This is done just to ensure any previously paused consumers get resumed on boot up.
consumer.Resume(ctx, messagebroker.ResumeOnTopicRequest{Topic: topicName})
return &DelayConsumer{
subscriberID: subscriberID,
ctx: delayCtx,
cancelFunc: cancel,
topic: topicName,
consumer: consumer,
subs: subs,
bs: bs,
handler: handler,
ch: ch,
doneCh: make(chan struct{}),
cachedMsgs: make([]messagebroker.ReceivedMessage, 0),
errChan: errChan,
}, nil
}
// Run spawns the delay-consumer
func (dc *DelayConsumer) Run(ctx context.Context) {
defer close(dc.doneCh)
logger.Ctx(ctx).Infow("delay-consumer: running", dc.LogFields()...)
for {
select {
case <-dc.ctx.Done():
logger.Ctx(dc.ctx).Infow("delay-consumer: stopping <-ctx.Done() called", dc.LogFields()...)
dc.bs.RemoveConsumer(ctx, messagebroker.ConsumerClientOptions{GroupID: dc.subs.GetDelayConsumerGroupID(dc.topic), GroupInstanceID: dc.subs.GetDelayConsumerGroupInstanceID(dc.subscriberID, dc.topic)})
dc.consumer.Close(dc.ctx)
return
default:
resp, err := dc.consumer.ReceiveMessages(dc.ctx, messagebroker.GetMessagesFromTopicRequest{NumOfMessages: 10, TimeoutMs: int(defaultBrokerOperationsTimeoutMs)})
if err != nil {
if !messagebroker.IsErrorRecoverable(err) {
logger.Ctx(dc.ctx).Errorw("delay-consumer: error in receiving messages", dc.LogFields("error", err.Error())...)
dc.errChan <- err
return
}
}
if len(resp.Messages) > 0 {
logger.Ctx(ctx).Infow("delay-consumer: non zero messages received", dc.LogFields("len", len(resp.Messages))...)
}
dc.cachedMsgs = append(dc.cachedMsgs, resp.Messages...)
dc.processMsgs()
}
}
}
// resumes a paused delay-consumer. Additionally process any previously cached messages in buffer.
func (dc *DelayConsumer) resume() {
logger.Ctx(dc.ctx).Infow("delay-consumer: resuming", dc.LogFields("error", nil)...)
dc.consumer.Resume(dc.ctx, messagebroker.ResumeOnTopicRequest{Topic: dc.topic})
dc.isPaused = false
}
// paused an active delay-consumer. Additionally caches the last seen message locally.
func (dc *DelayConsumer) pause() {
logger.Ctx(dc.ctx).Infow("delay-consumer: pausing", dc.LogFields()...)
dc.consumer.Pause(dc.ctx, messagebroker.PauseOnTopicRequest{Topic: dc.topic})
dc.isPaused = true
}
// push a message onto the configured dead letter topic
func (dc *DelayConsumer) pushToDeadLetter(msg *messagebroker.ReceivedMessage) error {
logger.Ctx(dc.ctx).Infow("delay-consumer: pushing to dead-letter", dc.LogFields()...)
dlProducer, err := dc.bs.GetProducer(dc.ctx, messagebroker.ProducerClientOptions{
Topic: msg.DeadLetterTopic,
TimeoutMs: defaultBrokerOperationsTimeoutMs,
})
if err != nil {
return err
}
msg.MessageHeader.ClosestDelayInterval = 0
msg.MessageHeader.CurrentDelayInterval = 0
msg.MessageHeader.CurrentRetryCount = 0
_, err = dlProducer.SendMessage(dc.ctx, messagebroker.SendMessageToTopicRequest{
Topic: msg.DeadLetterTopic,
Message: msg.Data,
OrderingKey: msg.OrderingKey,
TimeoutMs: int(defaultBrokerOperationsTimeoutMs),
MessageHeader: msg.MessageHeader,
Attributes: msg.Attributes,
})
if err != nil {
return err
}
subscriberTotalMessagesPushedToDLQ.WithLabelValues(env, msg.SourceTopic, msg.Subscription).Inc()
return nil
}
// processes a given message of the delay-topic. takes care of orchestrating the checks to determine pause,resume of the consumer,
// push to dead-letter topic in case the number of retries breaches allowed threshold.
func (dc *DelayConsumer) processMsgs() {
for len(dc.cachedMsgs) > 0 {
msg := dc.cachedMsgs[0]
if msg.CanProcessMessage() {
if dc.retryExhausted(msg) || (msg.CurrentRetryCount > msg.MaxRetryCount) {
// if the source topic is dlq-topic, message will be lost after exhausting max retries.
if !topic.IsDLQTopic(dc.subs.GetTopic()) {
// push to dead-letter topic directly in such cases
logger.Ctx(dc.ctx).Infow("delay-consumer: publishing to DLQ topic", dc.LogFields("messageID", msg.MessageID)...)
err := dc.pushToDeadLetter(&msg)
if err != nil {
logger.Ctx(dc.ctx).Errorw("delay-consumer: failed to push to dead-letter topic",
dc.LogFields("messageID", msg.MessageID, "topic", msg.DeadLetterTopic, "error", err.Error())...,
)
return
}
}
} else {
// submit to
logger.Ctx(dc.ctx).Infow("delay-consumer: processing message", dc.LogFields("messageID", msg.MessageID)...)
err := dc.incrementRetryCount(msg)
if err != nil {
logger.Ctx(dc.ctx).Errorw("delayconsumer: failed to increment retry count", err.Error())
}
err = dc.handler.Do(dc.ctx, msg)
if err != nil {
logger.Ctx(dc.ctx).Errorw("delay-consumer: error in msg handler",
dc.LogFields("messageID", msg.MessageID, "error", err.Error())...)
return
}
}
// commit the message
_, err := dc.consumer.CommitByPartitionAndOffset(dc.ctx, messagebroker.CommitOnTopicRequest{
Topic: msg.Topic,
Partition: msg.Partition,
Offset: msg.Offset + 1,
})
if err != nil {
logger.Ctx(dc.ctx).Errorw("delay-consumer: error on commit", dc.LogFields("error", err.Error())...)
return
}
// evict the processed message from cached messages
dc.cachedMsgs = dc.cachedMsgs[1:]
} else {
// pause the consumer, if not already paused
if !dc.isPaused {
logger.Ctx(dc.ctx).Infow("delay-consumer: pausing consumer for message",
dc.LogFields("messageID", msg.MessageID)...)
dc.pause()
}
break
}
}
if len(dc.cachedMsgs) == 0 && dc.isPaused {
// pull messages from the broker to process the next batch of messages
dc.resume()
}
}
// LogFields ...
func (dc *DelayConsumer) LogFields(kv ...interface{}) []interface{} {
fields := []interface{}{
"delayConsumerConfig", map[string]interface{}{
"topic": dc.topic,
"groupID": dc.subs.GetDelayConsumerGroupID(dc.topic),
"groupInstanceID": dc.subs.GetDelayConsumerGroupInstanceID(dc.subscriberID, dc.topic),
},
}
fields = append(fields, kv...)
return fields
}
func (dc *DelayConsumer) retryExhausted(msg messagebroker.ReceivedMessage) bool {
count, err := dc.fetchRetryCount(msg)
if err != nil {
logger.Ctx(dc.ctx).Errorw("delayconsumer: error fetching retry count from cache", "error", err.Error())
return false
}
if count >= int(msg.MaxRetryCount) {
return true
}
return false
}