consumer 消费失败,会把消息重新发往 %RETRY% + consumerGroup,这个 retry 消息会在一定时间后,真实送到 retry topic。

broker 处理发送到 retry topic 的消息:

org.apache.rocketmq.broker.processor.SendMessageProcessor#consumerSendMsgBack

消息消费超过最大次数或者客户端配置了直接发送到死信队列,则把消息发送到死信队列,否则把消息发送 retry topic,虽然看起来是把消息直接写入 %RETRY% + consumerGroup

但其实在 putMessage 的时候,会把消息写入 SCHEDULE_TOPIC_XXXX

// org.apache.rocketmq.store.CommitLog#putMessage
if (msg.getDelayTimeLevel() > 0) {
if (msg.getDelayTimeLevel() > this.defaultMessageStore.getScheduleMessageService().getMaxDelayLevel()) {
msg.setDelayTimeLevel(this.defaultMessageStore.getScheduleMessageService().getMaxDelayLevel());
}

topic = ScheduleMessageService.SCHEDULE_TOPIC;
queueId = ScheduleMessageService.delayLevel2QueueId(msg.getDelayTimeLevel());

// Backup real topic, queueId
MessageAccessor.putProperty(msg, MessageConst.PROPERTY_REAL_TOPIC, msg.getTopic());
MessageAccessor.putProperty(msg, MessageConst.PROPERTY_REAL_QUEUE_ID, String.valueOf(msg.getQueueId()));
msg.setPropertiesString(MessageDecoder.messageProperties2String(msg.getProperties()));

msg.setTopic(topic);
msg.setQueueId(queueId);
}

SCHEDULE_TOPIC_XXXX 这个 topic 非常有意思,broker 并没有显式创建这个 topic,即 nameserver 和 broker 没有保存这个 broker 的元数据,topic 的数据会正常写入 commitLog,一个 delay 等级对应一个 queue,queueId = delayLevel - 1,所以 SCHEDULE_TOPIC_XXXX 最多有 18 个 queue。

// org.apache.rocketmq.store.config.MessageStoreConfig#messageDelayLevel
private String messageDelayLevel = "1s 5s 10s 30s 1m 2m 3m 4m 5m 6m 7m 8m 9m 10m 20m 30m 1h 2h";

一共有 18 个 delayLevel

// org.apache.rocketmq.common.subscription.SubscriptionGroupConfig#retryMaxTimes
private int retryMaxTimes = 16;

这个参数 consumer 不可配置,默认 16

ScheduleMessageService 初始化 delayLevelTable,键是 delayLevel,值是 delay 的毫秒数,从 1 到 18

// org.apache.rocketmq.store.schedule.ScheduleMessageService#load
public boolean load() {
boolean result = super.load();
result = result && this.parseDelayLevel();
return result;
}
public boolean parseDelayLevel() {
HashMap<String, Long> timeUnitTable = new HashMap<String, Long>();
timeUnitTable.put("s", 1000L);
timeUnitTable.put("m", 1000L * 60);
timeUnitTable.put("h", 1000L * 60 * 60);
timeUnitTable.put("d", 1000L * 60 * 60 * 24);

String levelString = this.defaultMessageStore.getMessageStoreConfig().getMessageDelayLevel();
try {
String[] levelArray = levelString.split(" ");
for (int i = 0; i < levelArray.length; i++) {
String value = levelArray[i];
String ch = value.substring(value.length() - 1);
Long tu = timeUnitTable.get(ch);

int level = i + 1;
if (level > this.maxDelayLevel) {
this.maxDelayLevel = level;
}
long num = Long.parseLong(value.substring(0, value.length() - 1));
long delayTimeMillis = tu * num;
this.delayLevelTable.put(level, delayTimeMillis);
}
} catch (Exception e) {
log.error("parseDelayLevel exception", e);
log.info("levelString String = {}", levelString);
return false;
}

return true;
}

ScheduleMessageService 针对每一个 level 创建一个定时任务,遍历 consume queue,判断消息是否到期,到期则把消息写入真实 topic

// org.apache.rocketmq.store.schedule.ScheduleMessageService#start
public void start() {
for (Map.Entry<Integer, Long> entry : this.delayLevelTable.entrySet()) {
Integer level = entry.getKey();
Long timeDelay = entry.getValue();
Long offset = this.offsetTable.get(level);
if (null == offset) {
offset = 0L;
}

if (timeDelay != null) {
this.timer.schedule(new DeliverDelayedMessageTimerTask(level, offset), FIRST_DELAY_TIME);
}
}

// 定期持久化处理完的 queue offset 到 delayOffset.json 文件中
this.timer.scheduleAtFixedRate(new TimerTask() {

@Override
public void run() {
try {
ScheduleMessageService.this.persist();
} catch (Throwable e) {
log.error("scheduleAtFixedRate flush exception", e);
}
}
}, 10000, this.defaultMessageStore.getMessageStoreConfig().getFlushDelayOffsetInterval());
}