[vagrant@localhost bin]$ ./zkCli.sh
....
[zk: localhost:2181(CONNECTED) 1] ls /kafka
[admin, brokers, cluster, config, consumers, controller, controller_epoch, feature, isr_change_notification, latest_producer_id_block, log_dir_event_notification]
几个关键数据:
/kafka/brokers/ids记录有哪些服务器
[zk: localhost:2181(CONNECTED) 0] ls /kafka/brokers/ids
[0, 1, 2]
[zk: localhost:2181(CONNECTED) 1] get /kafka/brokers/ids/0
{"listener_security_protocol_map":{"PLAINTEXT":"PLAINTEXT"},"endpoints":["PLAINTEXT://192.168.3.51:9092"],"jmx_port":9999,"features":{},"host":"192.168.3.51","timestamp":"1647170725662","port":9092,"version":5}
[zk: localhost:2181(CONNECTED) 2] get /kafka/brokers/ids/1
{"listener_security_protocol_map":{"PLAINTEXT":"PLAINTEXT"},"endpoints":["PLAINTEXT://192.168.3.52:9092"],"jmx_port":9999,"features":{},"host":"192.168.3.52","timestamp":"1646125425746","port":9092,"version":5}
[zk: localhost:2181(CONNECTED) 4] get /kafka/brokers/ids/2
{"listener_security_protocol_map":{"PLAINTEXT":"PLAINTEXT"},"endpoints":["PLAINTEXT://192.168.3.53:9092"],"jmx_port":9999,"features":{},"host":"192.168.3.53","timestamp":"1646125296667","port":9092,"version":5}
/kafka/brokers/topics/first/partitions/0/state
记录谁是Leader,有哪些服务器可用
[zk: localhost:2181(CONNECTED) 7] ls /kafka/brokers/topics
[__consumer_offsets, first]
[zk: localhost:2181(CONNECTED) 17] ls /kafka/brokers/topics/first/partitions
[0, 1, 2]
[zk: localhost:2181(CONNECTED) 13] get /kafka/brokers/topics/first/partitions/0/state
{"controller_epoch":1,"leader":1,"version":1,"leader_epoch":0,"isr":[1]}
[zk: localhost:2181(CONNECTED) 14] get /kafka/brokers/topics/first/partitions/1/state
{"controller_epoch":1,"leader":2,"version":1,"leader_epoch":0,"isr":[2]}
[zk: localhost:2181(CONNECTED) 15] get /kafka/brokers/topics/first/partitions/2/state
{"controller_epoch":1,"leader":0,"version":1,"leader_epoch":0,"isr":[0]}
/kafka/controller
辅助选举Leader
[zk: localhost:2181(CONNECTED) 20] get /kafka/controller
{"version":1,"brokerid":0,"timestamp":"1647170725901"}
/kafka/consumers
0.9版本之前用于保存offset信息
0.9版本之后offset存储在kafka主题中,/kafka/brokers/topics/__consumer_offsets/partitions
参数名称 | 描述 |
---|---|
replica.lag.time.max.ms | ISR 中,如果 Follower 长时间未向 Leader 发送通信请求或同步数据,则该 Follower 将被踢出 ISR。 该时间阈值,默认 30s。 |
auto.leader.rebalance.enable | 默认是 true。 自动 Leader Partition 平衡。 |
leader.imbalance.per.broker.percentage | 默认是 10%。 每个 broker 允许的不平衡的 leader的比率。 如果每个 broker 超过了这个值,控制器会触发 leader 的平衡。 |
leader.imbalance.check.interval.seconds | 默认值 300 秒。 检查 leader 负载是否平衡的间隔时间。 |
log.segment.bytes | Kafka 中 log 日志是分成一块块存储的,此配置是指 log 日志划分 成块的大小,默认值 1G。 |
log.index.interval.bytes | 默认 4kb,kafka 里面每当写入了 4kb 大小的日志(.log),然后就往 index 文件里面记录一个索引。 |
log.retention.hours | Kafka 中数据保存的时间,默认 7 天。 |
log.retention.minutes | Kafka 中数据保存的时间,分钟级别,默认关闭。 |
log.retention.ms | Kafka 中数据保存的时间,毫秒级别,默认关闭。 |
log.retention.check.interval.ms | 检查数据是否保存超时的间隔,默认是 5 分钟。 |
log.retention.bytes | 默认等于-1,表示无穷大。 超过设置的所有日志总大小,删除最早的 segment。 |
log.cleanup.policy | 默认是 delete,表示所有数据启用删除策略;如果设置值为 compact,表示所有数据启用压缩策略。 |
num.io.threads | 默认是 8。 负责写磁盘的线程数。 整个参数值要占总核数的 50%。 |
num.replica.fetchers | 副本拉取线程数,这个参数占总核数的 50%的 1/3 |
num.network.threads | 默认是 3。 数据传输线程数,这个参数占总核数的50%的 2/3 。 |
log.flush.interval.messages | 强制页缓存刷写到磁盘的条数,默认是 long 的最大值,9223372036854775807。 一般不建议修改,交给系统自己管理。 |
log.flush.interval.ms | 每隔多久,刷数据到磁盘,默认是 null。 一般不建议修改,交给系统自己管理。 |
准备一台新机器:192.168.3.54
访问zk,查看first存储的信息
[vagrant@localhost bin]$ ./zkCli.sh
[zk: localhost:2181(CONNECTED) 2] get /kafka/config/topics/first
{"version":1,"config":{}}
查看first 主题的详细信息:
/home/vagrant/kafka_2.12-3.0.0/bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic first
Topic: first TopicId: zeecMSE5QSyIWnZWGxhJZQ PartitionCount: 3 ReplicationFactor: 1 Configs: segment.bytes=1073741824
Topic: first Partition: 0 Leader: 1 Replicas: 1 Isr: 1
Topic: first Partition: 1 Leader: 2 Replicas: 2 Isr: 2
Topic: first Partition: 2 Leader: 0 Replicas: 0 Isr: 0
服役新节点
2.1 创建一个要均衡的主题
vi topics-to-move.json
{
"topics": [
{
"topic": "first"
}
],
"version": 1
}
2.2 生成一个负载均衡的计划
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --topics-to-move-json-file topics-to-move.json --broker-list "0,1,2,3" --generate
Current partition replica assignment
{
"version": 1,
"partitions": [
{
"topic": "first",
"partition": 0,
"replicas": [
1
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 1,
"replicas": [
2
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 2,
"replicas": [
0
],
"log_dirs": [
"any"
]
}
]
}
Proposed partition reassignment configuration
{
"version": 1,
"partitions": [
{
"topic": "first",
"partition": 0,
"replicas": [
1
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 1,
"replicas": [
2
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 2,
"replicas": [
3
],
"log_dirs": [
"any"
]
}
]
}
2.3 创建副本存储计划(所有副本存储在 broker0、broker1、broker2、broker3 中)
vi increase-replication-factor.json
{
"version": 1,
"partitions": [
{
"topic": "first",
"partition": 0,
"replicas": [
1
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 1,
"replicas": [
2
],
"log_dirs": [
"any"
]
},
{
"topic": "first",
"partition": 2,
"replicas": [
3
],
"log_dirs": [
"any"
]
}
]
}
2.4 执行副本存储计划
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --reassignment-json-file increase-replication-factor.json --execute
Current partition replica assignment
{"version":1,"partitions":[{"topic":"first","partition":0,"replicas":[1],"log_dirs":["any"]},{"topic":"first","partition":1,"replicas":[2],"log_dirs":["any"]},{"topic":"first","partition":2,"replicas":[0],"log_dirs":["any"]}]}
Save this to use as the --reassignment-json-file option during rollback
Successfully started partition reassignments for first-0,first-1,first-2
2.5 验证副本存储计划
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --reassignment-json-file increase-replication-factor.json --verify
Status of partition reassignment:
Reassignment of partition first-0 is complete.
Reassignment of partition first-1 is complete.
Reassignment of partition first-2 is complete.
Clearing broker-level throttles on brokers 0,1,2,3
Clearing topic-level throttles on topic first
再次查看first 主题的详细信息:
/home/vagrant/kafka_2.12-3.0.0/bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic first
Topic: first TopicId: zeecMSE5QSyIWnZWGxhJZQ PartitionCount: 3 ReplicationFactor: 1 Configs: segment.bytes=1073741824
Topic: first Partition: 0 Leader: 1 Replicas: 1 Isr: 1
Topic: first Partition: 1 Leader: 2 Replicas: 2 Isr: 2
Topic: first Partition: 2 Leader: 3 Replicas: 3 Isr: 3
退役旧节点
先按照退役一台节点,生成执行计划,然后按照服役时 *** 作流程执行负载均衡。
类似服役新节点 *** 作,此处不做赘述。
- Kafka 副本作用:提高数据可靠性。
- Kafka 默认副本 1 个,生产环境一般配置为 2 个,保证数据可靠性;太多副本会增加磁盘存储空间,增加网络上数据传输,降低效率。
- Kafka 中副本分为:Leader 和 Follower。
Kafka 生产者只会把数据发往 Leader,然后 Follower 找 Leader 进行同步数据。
- Kafka 分区中的所有副本统称为 AR(Assigned Repllicas)。
AR = ISR + OSR
ISR,表示和 Leader 保持同步的 Follower 集合。如果 Follower 长时间未向 Leader 发送通信请求或同步数据,则该 Follower 将被踢出ISR。
该时间阈值由
replica.lag.time.max.ms
参数设定,默认 30s。Leader 发生故障之后,就会从 ISR 中选举新的 Leader。
参数设定,默认 30s。
Leader 发生故障之后,就会从 ISR 中选举新的 Leader。
OSR,表示 Follower 与 Leader 副本同步时,延迟过多的副本。,表示 Follower 与 Leader 副本同步时,延迟过多的副本。
Kafka 集群中有一个 broker 的 Controller 会被选举为 Controller Leader,负责管理集群broker 的上下线,所有 topic 的分区副本分配和 Leader 选举等工作。
Controller 的信息同步工作是依赖于 Zookeeper 的。
- 创建一个新的 topic,4 个分区,4 个副本
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --create --topic second --partitions 4 --replication-factor 4
Created topic second.
- 查看 Leader 分布情况
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic second
Topic: second TopicId: -DB-j_GCR0qOiVOkMGdzLg PartitionCount: 4 ReplicationFactor: 4 Configs: segment.bytes=1073741824
Topic: second Partition: 0 Leader: 0 Replicas: 0,3,1,2 Isr: 0,3,1,2
Topic: second Partition: 1 Leader: 2 Replicas: 2,1,0,3 Isr: 2,1,0,3
Topic: second Partition: 2 Leader: 3 Replicas: 3,0,2,1 Isr: 3,0,2,1
Topic: second Partition: 3 Leader: 1 Replicas: 1,2,3,0 Isr: 1,2,3,0
- 停止掉 192.168.3.54(id=3) 的 kafka 进程,并查看 Leader 分区情况
bin/kafka-server-stop.sh
[vagrant@localhost kafka_2.12-3.0.0]$ bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic second
[2022-03-24 14:02:28,273] WARN [AdminClient clientId=adminclient-1] Connection to node -4 (/192.168.3.54:9092) could not be established. Broker may not be available. (org.apache.kafka.clients.NetworkClient)
Topic: second TopicId: -DB-j_GCR0qOiVOkMGdzLg PartitionCount: 4 ReplicationFactor: 4 Configs: segment.bytes=1073741824
Topic: second Partition: 0 Leader: 0 Replicas: 0,3,1,2 Isr: 0,1,2
Topic: second Partition: 1 Leader: 2 Replicas: 2,1,0,3 Isr: 2,1,0
Topic: second Partition: 2 Leader: 0 Replicas: 3,0,2,1 Isr: 0,2,1
Topic: second Partition: 3 Leader: 1 Replicas: 1,2,3,0 Isr: 1,2,0
3.3 Leader 和 Follower 故障处理细节
LEO(Log End Offset):每个副本的最后一个offset,LEO其实就是最新的offset + 1。
HW(High Watermark):所有副本中最小的LEO 。
Follower发生故障后会被临时踢出ISR,这个期间Leader和Follower继续接收数据;待该Follower恢复后,Follower会读取本地磁盘记录的
上次的HW,并将log文件高于HW的部分截取掉,从HW开始向Leader进行同步。
等该Follower的LEO大于等于该Partition的HW,即
Follower追上Leader之后,就可以重新加入ISR了。
Leader发生故障之后,会从ISR中选出一个新的Leader,为保证多个副本之间的数据一致性,其余的Follower会先
将各自的log文件高于HW的部分截掉,然后从新的Leader同步数据。
注意:这只能保证副本之间的数据一致性,并不能保证数据不丢失或者不重复。
如果 kafka 服务器只有 4 个节点,那么设置 kafka 的分区数大于服务器台数,在 kafka底层如何分配存储副本呢?
1)创建 16 分区,3 个副本
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --create --partitions 16 --replication-factor 3 --topic third
Created topic third.
查看分区和副本情况。
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic third
Topic: third TopicId: RUzXbc22ThiH6usi2vux5Q PartitionCount: 16 ReplicationFactor: 3 Configs: segment.bytes=1073741824
Topic: third Partition: 0 Leader: 1 Replicas: 1,0,2 Isr: 1,0,2
Topic: third Partition: 1 Leader: 0 Replicas: 0,2,3 Isr: 0,2,3
Topic: third Partition: 2 Leader: 2 Replicas: 2,3,1 Isr: 2,3,1
Topic: third Partition: 3 Leader: 3 Replicas: 3,1,0 Isr: 3,1,0
Topic: third Partition: 4 Leader: 1 Replicas: 1,2,3 Isr: 1,2,3
Topic: third Partition: 5 Leader: 0 Replicas: 0,3,1 Isr: 0,3,1
Topic: third Partition: 6 Leader: 2 Replicas: 2,1,0 Isr: 2,1,0
Topic: third Partition: 7 Leader: 3 Replicas: 3,0,2 Isr: 3,0,2
Topic: third Partition: 8 Leader: 1 Replicas: 1,3,0 Isr: 1,3,0
Topic: third Partition: 9 Leader: 0 Replicas: 0,1,2 Isr: 0,1,2
Topic: third Partition: 10 Leader: 2 Replicas: 2,0,3 Isr: 2,0,3
Topic: third Partition: 11 Leader: 3 Replicas: 3,2,1 Isr: 3,2,1
Topic: third Partition: 12 Leader: 1 Replicas: 1,0,2 Isr: 1,0,2
Topic: third Partition: 13 Leader: 0 Replicas: 0,2,3 Isr: 0,2,3
Topic: third Partition: 14 Leader: 2 Replicas: 2,3,1 Isr: 2,3,1
Topic: third Partition: 15 Leader: 3 Replicas: 3,1,0 Isr: 3,1,0
3.5 手动调整分区副本存储
在生产环境中,每台服务器的配置和性能不一致,但是Kafka只会根据自己的代码规则创建对应的分区副本,就会导致个别服务器存储压力较大。
所有需要手动调整分区副本的存储。
需求:创建一个新的topic,4个分区,两个副本,名称为three。
将 该topic的所有副本都存储到broker0和broker1两台服务器上。
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --create --partitions 4 --replication-factor 2 --topic three
Created topic three.
查看分区副本存储情况
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic three
Topic: three TopicId: DWrKh6rKRJ6Y2NEtKgCkOw PartitionCount: 4 ReplicationFactor: 2 Configs: segment.bytes=1073741824
Topic: three Partition: 0 Leader: 3 Replicas: 3,1 Isr: 3,1
Topic: three Partition: 1 Leader: 1 Replicas: 1,0 Isr: 1,0
Topic: three Partition: 2 Leader: 0 Replicas: 0,2 Isr: 0,2
Topic: three Partition: 3 Leader: 2 Replicas: 2,3 Isr: 2,3
创建副本存储计划(所有副本都指定存储在 broker0、broker1 中)
vi increase-replication-factor.json
{
"version": 1,
"partitions": [
{
"topic": "three",
"partition": 0,
"replicas": [
0,
1
]
},
{
"topic": "three",
"partition": 1,
"replicas": [
0,
1
]
},
{
"topic": "three",
"partition": 2,
"replicas": [
1,
0
]
},
{
"topic": "three",
"partition": 3,
"replicas": [
1,
0
]
}
]
}
执行副本存储计划
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --reassignment-json-file increase-replication-factor.json --execute
Current partition replica assignment
{"version":1,"partitions":[{"topic":"three","partition":0,"replicas":[3,1],"log_dirs":["any","any"]},{"topic":"three","partition":1,"replicas":[1,0],"log_dirs":["any","any"]},{"topic":"three","partition":2,"replicas":[0,2],"log_dirs":["any","any"]},{"topic":"three","partition":3,"replicas":[2,3],"log_dirs":["any","any"]}]}
Save this to use as the --reassignment-json-file option during rollback
Successfully started partition reassignments for three-0,three-1,three-2,three-3
验证副本存储计划
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --reassignment-json-file increase-replication-factor.json --verify
Status of partition reassignment:
Reassignment of partition three-0 is complete.
Reassignment of partition three-1 is complete.
Reassignment of partition three-2 is complete.
Reassignment of partition three-3 is complete.
Clearing broker-level throttles on brokers 0,1,2,3
Clearing topic-level throttles on topic three
查看分区副本存储情况
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic three
Topic: three TopicId: DWrKh6rKRJ6Y2NEtKgCkOw PartitionCount: 4 ReplicationFactor: 2 Configs: segment.bytes=1073741824
Topic: three Partition: 0 Leader: 0 Replicas: 0,1 Isr: 1,0
Topic: three Partition: 1 Leader: 1 Replicas: 0,1 Isr: 1,0
Topic: three Partition: 2 Leader: 0 Replicas: 1,0 Isr: 0,1
Topic: three Partition: 3 Leader: 1 Replicas: 1,0 Isr: 0,1
3.6 Leader Partition 负载平衡
正常情况下,Kafka本身会自动把Leader Partition均匀分散在各个机器上,来保证每台机器的读写吞吐量都是均匀的。
但是如果某 些broker宕机,会导致Leader Partition过于集中在其他少部分几台broker上,这会导致少数几台broker的读写请求压力过高,其他宕机的broker重启之后都是follower partition,读写请求很低,造成集群负载不均衡。
auto.leader.rebalance.enable
,默认是true。自动Leader Partition 平衡
leader.imbalance.per.broker.percentage
,默认是10%。每个broker允许的不平衡的leader的比率。
如果每个broker超过了这个值,控制器会触发leader的平衡。
leader.imbalance.check.interval.seconds
,默认值300秒。检查leader负载是否平衡的间隔时间。
下面拿一个主题举例说明,假设集群只有一个主题如下图所示:
Topic: second TopicId: -DB-j_GCR0qOiVOkMGdzLg PartitionCount: 4 ReplicationFactor: 4 Configs: segment.bytes=1073741824
Topic: second Partition: 0 Leader: 0 Replicas: 1,0,3,2 Isr: 0,3,1,2
Topic: second Partition: 1 Leader: 2 Replicas: 2,1,0,3 Isr: 2,1,0,3
Topic: second Partition: 2 Leader: 3 Replicas: 0,3,2,1 Isr: 3,0,2,1
Topic: second Partition: 3 Leader: 1 Replicas: 3,2,1,0 Isr: 1,2,3,0
针对broker0节点,分区2的AR优先副本是0节点,但是0节点却不是Leader节点,所以不平衡数加1,AR副本总数是4,所以broker0节点不平衡率为1/4>10%,需要再平衡。
broker2和broker3节点和broker0不平衡率一样,需要再平衡。
Broker1的不平衡数为0,不需要再平衡。
参数名称 | 描述 |
---|---|
auto.leader.rebalance.enable | 默认是 true。 自动 Leader Partition 平衡。 生产环境中,leader 重选举的代价比较大,可能会带来性能影响,建议设置为 false 关闭。 |
leader.imbalance.per.broker.percentage | 默认是 10%。 每个 broker 允许的不平衡的 leader的比率。 如果每个 broker 超过了这个值,控制器会触发 leader 的平衡。 |
leader.imbalance.check.interval.seconds | 默认值 300 秒。 检查 leader 负载是否平衡的间隔时间。 |
在生产环境当中,由于某个主题的重要等级需要提升,我们考虑增加副本。
副本数的增加需要先制定计划,然后根据计划执行。
创建 topic
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --create --partitions 3 --replication-factor 1 --topic four
Created topic four.
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic four
Topic: four TopicId: 3BdiSOnBSjiM9gXsHEpfpQ PartitionCount: 3 ReplicationFactor: 1 Configs: segment.bytes=1073741824
Topic: four Partition: 0 Leader: 0 Replicas: 0 Isr: 0
Topic: four Partition: 1 Leader: 2 Replicas: 2 Isr: 2
Topic: four Partition: 2 Leader: 3 Replicas: 3 Isr: 3
手动增加副本存储
(1)创建副本存储计划(所有副本都指定存储在 broker0、broker1、broker2 中)。
vi increase-replication-factor.json
{
"version": 1,
"partitions": [
{
"topic": "four",
"partition": 0,
"replicas": [
0,
1,
2
]
},
{
"topic": "four",
"partition": 1,
"replicas": [
0,
1,
2
]
},
{
"topic": "four",
"partition": 2,
"replicas": [
0,
1,
2
]
}
]
}
(2)执行副本存储计划。
bin/kafka-reassign-partitions.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --reassignment-json-file increase-replication-factor.json --execute
Current partition replica assignment
{"version":1,"partitions":[{"topic":"four","partition":0,"replicas":[0],"log_dirs":["any"]},{"topic":"four","partition":1,"replicas":[2],"log_dirs":["any"]},{"topic":"four","partition":2,"replicas":[3],"log_dirs":["any"]}]}
Save this to use as the --reassignment-json-file option during rollback
Successfully started partition reassignments for four-0,four-1,four-2
再次查看
bin/kafka-topics.sh --bootstrap-server 192.168.3.51:9092,192.168.3.52:9092,192.168.3.53:9092,192.168.3.54:9092 --describe --topic four
Topic: four TopicId: 3BdiSOnBSjiM9gXsHEpfpQ PartitionCount: 3 ReplicationFactor: 3 Configs: segment.bytes=1073741824
Topic: four Partition: 0 Leader: 0 Replicas: 0,1,2 Isr: 0,2,1
Topic: four Partition: 1 Leader: 2 Replicas: 0,1,2 Isr: 2,0,1
Topic: four Partition: 2 Leader: 0 Replicas: 0,1,2 Isr: 0,2,1
4 文件存储
4.1 文件存储机制
Topic 数据的存储机制
Topic是逻辑上的概念,而partition是物理上的概念,每个partition对应于一个log文件,该log文件中存储的就是Producer生产的数据。
Producer生产的数据会被不断追加到该log文件末端,为防止log文件过大导致数据定位效率低下,Kafka采取了分片和索引机制, 将每个partition分为多个segment。
每个segment包括:“.index”文件、“.log”文件和.timeindex等文件。
这些文件位于一个文件夹下,该文件夹的命名规则为:topic名称+分区序号,例如:first-0。
启动生产者,并发送消息
# 创建一个新主题five
> bin/kafka-topics.sh --bootstrap-server 192.168.3.52:9092 --create --partitions 1 --replication-factor 1 --topic five
# 启动一个生产者
> bin/kafka-console-producer.sh --bootstrap-server 192.168.3.52:9092 --topic five
>hell01
>hello2
>hello3
>hello4
>hello5
> cd /home/vagrant/kafka_2.12-3.0.0/data/five-0
> ll
total 12
-rw-r--r--. 1 root root 10485760 Mar 15 06:13 00000000000000000000.index
-rw-r--r--. 1 root root 222 Mar 15 06:26 00000000000000000000.log
-rw-r--r--. 1 root root 10485756 Mar 15 06:13 00000000000000000000.timeindex
-rw-r--r--. 1 root root 8 Mar 15 06:13 leader-epoch-checkpoint
-rw-r--r--. 1 root root 43 Mar 15 06:13 partition.metadata
# 查看 index 日志
> kafka-run-class.sh kafka.tools.DumpLogSegments --files 00000000000000000005.index
Dumping 00000000000000000005.index
offset: 63 position: 4102
offset: 118 position: 8246
# 查看 log
> kafka-run-class.sh kafka.tools.DumpLogSegments --files ./00000000000000000005.log
Dumping 00000000000000000005.log
Starting offset: 5
baseOffset: 5 lastOffset: 5 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 0 CreateTime: 1648298105592 size: 74 magic: 2 compresscodec: none crc: 2922802256 isvalid: true
baseOffset: 6 lastOffset: 7 count: 2 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 74 CreateTime: 1648298105628 size: 87 magic: 2 compresscodec: none crc: 2510574599 isvalid: true
baseOffset: 8 lastOffset: 11 count: 4 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 161 CreateTime: 1648298105639 size: 116 magic: 2 compresscodec: none crc: 2793496984 isvalid: true
baseOffset: 12 lastOffset: 12 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 277 CreateTime: 1648298105642 size: 75 magic: 2 compresscodec: none crc: 1332889768 isvalid: true
baseOffset: 13 lastOffset: 13 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 352 CreateTime: 1648298105645 size: 75 magic: 2 compresscodec: none crc: 2515862546 isvalid: true
baseOffset: 14 lastOffset: 14 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 427 CreateTime: 1648298105648 size: 75 magic: 2 compresscodec: none crc: 3992299684 isvalid: true
baseOffset: 15 lastOffset: 15 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 502 CreateTime: 1648298105651 size: 75 magic: 2 compresscodec: none crc: 4170130861 isvalid: true
baseOffset: 16 lastOffset: 16 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 577 CreateTime: 1648298105654 size: 75 magic: 2 compresscodec: none crc: 2307576682 isvalid: true
baseOffset: 17 lastOffset: 17 count: 1 baseSequence: -1 lastSequence: -1 producerId: -1 producerEpoch: -1 partitionLeaderEpoch: 0 isTransactional: false isControl: false position: 652 CreateTime: 1648298105657 size: 75 magic: 2 compresscodec: none crc: 3373951559 isvalid: true
Log文件和Index文件详解
说明:日志存储参数配置
参数 | 描述 |
---|---|
log.segment.bytes | Kafka 中 log 日志是分成一块块存储的,此配置是指 log 日志划分成块的大小,默认值 1G。 |
log.index.interval.bytes | 默认 4kb,kafka 里面每当写入了 4kb 大小的日志(.log),然后就往 index 文件里面记录一个索引。 稀疏索引。 |
Kafka 中默认的日志保存时间为 7 天,可以通过调整如下参数修改保存时间。
- log.retention.hours,最低优先级小时,默认 7 天。
- log.retention.minutes,分钟。
- log.retention.ms,最高优先级毫秒。
- log.retention.check.interval.ms,负责设置检查周期,默认 5 分钟。
那么日志一旦超过了设置的时间,怎么处理呢?
Kafka 中提供的日志清理策略有 delete 和 compact 两种。
1)delete 日志删除:将过期数据删除
- log.cleanup.policy = delete 所有数据启用删除策略
(1)基于时间:默认打开。以 segment 中所有记录中的最大时间戳作为该文件时间戳。
(2)基于大小:默认关闭。超过设置的所有日志总大小,删除最早的 segment。
log.retention.bytes,默认等于-1,表示无穷大。
思考:如果一个 segment 中有一部分数据过期,一部分没有过期,怎么处理?
以 segment 中所有记录中的最大时间戳作为该文件时间戳,最大的时间戳过期整个文件才过期。
2)compact 日志压缩
compact日志压缩:对于相同key的不同value值,只保留最后一个版本。
- log.cleanup.policy = compact 所有数据启用压缩策略
压缩后的offset可能是不连续的,比如上图中没有6,当从这些offset消费消息时,将会拿到比这个offset大 的offset对应的消息,实际上会拿到offset为7的消息,并从这个位置开始消费。
这种策略只适合特殊场景,比如消息的key是用户ID,value是用户的资料,通过这种压缩策略,整个消息集里就保存了所有用户最新的资料。
- Kafka 本身是分布式集群,可以采用分区技术,并行度高
- 读数据采用稀疏索引,可以快速定位要消费的数据
- 顺序写磁盘
Kafka 的 producer 生产数据,要写入到 log 文件中,写的过程是一直追加到文件末端,为顺序写。官网有数据表明,同样的磁盘,顺序写能到 600M/s,而随机写只有 100K/s。
这与磁盘的机械机构有关,顺序写之所以快,是因为其省去了大量磁头寻址的时间。
- 页缓存 + 零拷贝技术
零拷贝:Kafka的数据加工处理 *** 作交由Kafka生产者和Kafka消费者处理。Kafka Broker应用层不关心存储的数据,所以就不用走应用层,传输效率高。
PageCache页缓存:Kafka重度依赖底层 *** 作系统提供的PageCache功 能。当上层有写 *** 作时, *** 作系统只是将数据写入PageCache。
当读 *** 作发生时,先从PageCache中查找,如果找不到,再去磁盘中读取。
实际上PageCache是把尽可能多的空闲内存都当做了磁盘缓存来使用。
参数 | 描述 |
---|---|
log.flush.interval.messages | 强制页缓存刷写到磁盘的条数,默认是 long 的最大值,9223372036854775807。 一般不建议修改,交给系统自己管理。 |
log.flush.interval.ms | 每隔多久,刷数据到磁盘,默认是 null。 一般不建议修改,交给系统自己管理。 |
欢迎分享,转载请注明来源:内存溢出
评论列表(0条)