ConsumeFromWhere
https://blog.csdn.net/prestigeding/article/details/96576932
对于一个新的消费组,无论是集群模式还是广播模式都不会存储该消费组的消费进度,可以理解为-1
key | remark |
---|---|
CONSUME_FROM_LAST_OFFSET | 默认策略,从该队列最尾开始消费,即跳过历史消息 |
CONSUME_FROM_FIRST_OFFSET | 从队列最开始开始消费,即设置从最小的 Offset开始读取 |
CONSUME_FROM_TIMESTAMP | 从某个时间点开始消费,和setConsumeTimestamp()配合使用,默认为消费者启动之前的30分钟处开始消费 |
注意设置读取位置不是每次都有效,它的优先级默认在 Offset Store后面 , 比如在 DefaultMQPushConsumer 的 BROADCASTING 方式下 ,默认是从 Broker 里读取某个 Topic 对应 ConsumerGroup 的 Offset, 当读取不到 Offset 的时候, ConsumeFromWhere 的设置才生效 。
大部分情况下这个设置在 Consumer Group初次启动时有效。 如果 Consumer正常运行后被停止, 然后再启动, 会 接着上次的 Offset开始消费, ConsumeFromWhere 的设置元效。
消息偏移量 Offset
https://www.cnblogs.com/jwen1994/p/12369913.html
/**
* Offset store interface
*/
public interface OffsetStore {
/**
* Load
*/
// LocalFileOffsetStore 实现 MixAll.file2String(this.storePath) || this.readLocalOffsetBak()
// RemoteBrokerOffsetStore 实现为空方法
void load() throws MQClientException;
/**
* Update the offset,store it in memory
*/
void updateOffset(final MessageQueue mq, final long offset, final boolean increaseOnly);
/**
* Get offset from local storage
* @return The fetched offset
*/
long readOffset(final MessageQueue mq, final ReadOffsetType type);
/**
* Persist all offsets,may be in local storage or remote name server
*/
void persistAll(final Set<MessageQueue> mqs);
/**
* Persist the offset,may be in local storage or remote name server
*/
void persist(final MessageQueue mq);
/**
* Remove offset
*/
void removeOffset(MessageQueue mq);
/**
* @return The cloned offset table of given topic
*/
Map<MessageQueue, Long> cloneOffsetTable(String topic);
/**
* @param mq
* @param offset
* @param isOneway
*/
// LocalFileOffsetStore 的实现为空方法
// RemoteBrokerOffsetStore 同步offset到Broker
/**
* Update the Consumer Offset synchronously, once the Master is off, updated to Slave, here need to be optimized.
*/
void updateConsumeOffsetToBroker(MessageQueue mq, long offset, boolean isOneway) throws RemotingException,
MQBrokerException, InterruptedException, MQClientException;
}
DefaultMQPushConsumer的BROADCASTING模式,各个Consumer没有互相干扰,使用LocalFileOffsetStore把Offset存储在本地
// package org.apache.rocketmq.client.consumer.store;
public class LocalFileOffsetStore implements OffsetStore {
public final static String LOCAL_OFFSET_STORE_DIR = System.getProperty(
"rocketmq.client.localOffsetStoreDir",
System.getProperty("user.home") + File.separator + ".rocketmq_offsets");
private final static InternalLogger log = ClientLogger.getLog();
private final MQClientInstance mQClientFactory;
private final String groupName;
private final String storePath;
private ConcurrentMap<MessageQueue, AtomicLong> offsetTable =
new ConcurrentHashMap<MessageQueue, AtomicLong>();
// ...
DefaultMQPushConsumer的CLUSTERING模式,由Broker端存储和控制Offset的值,使用RemoteBrokerOffsetStore
// package org.apache.rocketmq.client.consumer.store;
public class RemoteBrokerOffsetStore implements OffsetStore {
private final static InternalLogger log = ClientLogger.getLog();
private final MQClientInstance mQClientFactory;
private final String groupName;
private ConcurrentMap<MessageQueue, AtomicLong> offsetTable =
new ConcurrentHashMap<MessageQueue, AtomicLong>();
//...
}
建议采用pushConsumer,RocketMQ自动维护OffsetStore,如果用另外一种pullConsumer需要自己进行维护OffsetStore
// this.defaultLitePullConsumerImpl.start()
private void initOffsetStore() throws MQClientException {
if (this.defaultLitePullConsumer.getOffsetStore() != null) {
this.offsetStore = this.defaultLitePullConsumer.getOffsetStore();
} else {
switch (this.defaultLitePullConsumer.getMessageModel()) {
case BROADCASTING:
this.offsetStore = new LocalFileOffsetStore(this.mQClientFactory, this.defaultLitePullConsumer.getConsumerGroup());
break;
case CLUSTERING:
this.offsetStore = new RemoteBrokerOffsetStore(this.mQClientFactory, this.defaultLitePullConsumer.getConsumerGroup());
break;
default:
break;
}
this.defaultLitePullConsumer.setOffsetStore(this.offsetStore);
}
this.offsetStore.load();
}
message queue 是无限长的数组,一条消息进来下标就会涨1,下标就是 offset,消息在某个 MessageQueue 里的位置,通过 offset 的值可以定位到这条消息,或者指示 Consumer 从这条消息开始向后处理。
message queue 中的 maxOffset 表示消息的最大 offset,maxOffset 并不是最新的那条消息的 offset,而是最新消息的 offset+1,minOffset 则是现存在的最小 offset。
fileReserveTime=48 默认消息存储48小时后,消费会被物理地从磁盘删除,message queue 的minOffset 也就对应增长。所以比 minOffset 还要小的那些消息已经不在 broker上了,就无法被消费