Kafka 业务日志采集最佳实践

简介

Apache Kafka 是一个分布式流处理平台,主要用于构建实时数据流管道和应用程序。在收集业务日志的场景中,Kafka 可以作为一个消息中间件,用于接收、存储和转发大量的日志数据。将 Kafka 与其他系统(如 Elasticsearch、Flume、Spark Streaming 等)集成,以提供更丰富的日志处理和分析功能。本文提到的是和观测云集成,即通过观测云的采集器 Datakit 采集 Kafka 中的业务日志,下面通过一些例子了解下观测云的快速集成效果。

实践环境

前置条件

软件和中间件

  • Kafka3.2.0
  • Datakit采集器
  • JDK 8

硬件

  • 云服务器 CentOS7.9 64位 4vCPU,8GB 内存,100GB 云盘一台。

接入方案

准备 Kafka 环境

安装 Kafka

下载 3.2.0 版本,解压即可使用。

wget https://archive.apache.org/dist/kafka/3.2.0/kafka_2.13-3.2.0.tgz

注:目前 Datakit 支持的 Kafka 版本有[version:0.8.2 ~ 3.2.0]

启动 Zookeeper 服务
$ bin/zookeeper-server-start.sh config/zookeeper.properties
启动 KafkaServer
$ bin/kafka-server-start.sh config/server.properties
创建 Topic

创建名为 testlog 的 Topic 。

$ bin/kafka-topics.sh --create --topic testlog --bootstrap-server localhost:9092
启动 Producer
$ bin/kafka-console-producer.sh --topic testlog --bootstrap-server localhost:9092

安装 DataKit

参考官网文档安装 DataKit 采集器

TOKEN 依据你的观测云工作空间来填写
DK_DATAWAY=https://openway.guance.com?token=<TOKEN> bash -c "$(curl -L https://static.guance.com/datakit/install.sh)"

开启 Kafka 采集器

进入 DataKit 安装目录下 (默认是 /usr/local/datakit/conf.d/ ) 的 conf.d/kafkamq 目录,复制 kafkamq.conf.sample 并命名为 kafkamq.conf

类似如下:

-rwxr-xr-x 1 root root 2574 Apr 30 23:52 kafkamq.conf
-rwxr-xr-x 1 root root 2579 May  1 00:40 kafkamq.conf.sample

调制 kafka 采集器配置如下:

  • addrs = ["localhost:9092"],该文采集器 DataKit 和 Kafka 安装到同一台操作系统中,localhost 即可。

  • kafka_version = "3.2.0",该文使用 Kafka 的版本。

  • [inputs.kafkamq.custom],删除注释符号"#"。

  • [inputs.kafkamq.custom.log_topic_map],删除注释符号"#"。

  • "testlog"="log.p",testlog 为 Topic 的名字,log.p 为观测云 Pipeline 可编程数据处理器的日志字段提取规则配置。涉及的业务日志和 log.p 的内容详细见下面的《使用 Pipeline》。

    {"version": "1.28.1", "desc": "do NOT edit this line"}

    [[inputs.kafkamq]]
    addrs = ["localhost:9092"]
    # your kafka version:0.8.2 ~ 3.2.0
    kafka_version = "3.2.0"
    group_id = "datakit-group"
    # consumer group partition assignment strategy (range, roundrobin, sticky)
    assignor = "roundrobin"

    ## rate limit.
    #limit_sec = 100
    ## sample
    # sampling_rate = 1.0
    
    ## kafka tls config
    # tls_enable = true
    # tls_security_protocol = "SASL_PLAINTEXT"
    # tls_sasl_mechanism = "PLAIN"
    # tls_sasl_plain_username = "user"
    # tls_sasl_plain_password = "pw"
    
    ## -1:Offset Newest, -2:Offset Oldest
    offsets=-1
    
    ## skywalking custom
    #[inputs.kafkamq.skywalking]
      ## Required!send to datakit skywalking input.
      #dk_endpoint="http://localhost:9529"
      #thread = 8 
      #topics = [
      #  "skywalking-metrics",
      #  "skywalking-profilings",
      #  "skywalking-segments",
      #  "skywalking-managements",
      #  "skywalking-meters",
      #  "skywalking-logging",
      #]
      #namespace = ""
    
    ## Jaeger from kafka. Please make sure your Datakit Jaeger collector is open !!!
    #[inputs.kafkamq.jaeger]
      ## Required! ipv6 is "[::1]:9529"
      #dk_endpoint="http://localhost:9529"
      #thread = 8 
      #source: agent,otel,others...
      #source = "agent"
      ## Required! topics
      #topics=["jaeger-spans","jaeger-my-spans"]
    
    ## user custom message with PL script.
    [inputs.kafkamq.custom]
      #spilt_json_body = true
      #thread = 8 
      ## spilt_topic_map determines whether to enable log splitting for specific topic based on the values in the spilt_topic_map[topic].
      #[inputs.kafkamq.custom.spilt_topic_map]
      #  "log_topic"=true
      #  "log01"=false
      [inputs.kafkamq.custom.log_topic_map]
        "testlog"="log.p"
      #  "log01"="log_01.p"
      #[inputs.kafkamq.custom.metric_topic_map]
      #  "metric_topic"="metric.p"
      #  "metric01"="rum_apm.p"
      #[inputs.kafkamq.custom.rum_topic_map]
      #  "rum_topic"="rum_01.p"
      #  "rum_02"="rum_02.p"
    
    #[inputs.kafkamq.remote_handle]
      ## Required!
      #endpoint="http://localhost:8080"
      ## Required! topics
      #topics=["spans","my-spans"]
      # send_message_count = 100
      # debug = false
      # is_response_point = true
      # header_check = false
    
    ## Receive and consume OTEL data from kafka.
    #[inputs.kafkamq.otel]
      #dk_endpoint="http://localhost:9529"
      #trace_api="/otel/v1/trace"
      #metric_api="/otel/v1/metric"
      #trace_topics=["trace1","trace2"]
      #metric_topics=["otel-metric","otel-metric1"]
      #thread = 8 
    
    ## todo: add other input-mq
    

注意:开启或调整 DataKit 的配置,需重启采集器(shell 下执行 datakit service -R)。

使用 Pipeline

log.p 规则内容

data = load_json(message)
protocol = data["protocol"]
response_code = data["response_code"]
set_tag(protocol,protocol)
set_tag(response_code,response_code)
group_between(response_code,[200,300],"info","status")
group_between(response_code,[400,499],"warning","status")
group_between(response_code,[500,599],"error","status")

time = data["start_time"]
set_tag(time,time)
default_time(time)

效果展示

发送业务日志样例

业务日志样例文件如下:

#info
{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":204,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-05-01T00:37:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}
#error
{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":504,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-05-01T00:39:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}
#warn
{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":404,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-05-01T00:38:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}

日志发送命令

在 Producer 启动后,分别发送如下三条日志内容,三条日志一条为 info 级别("response_code":204),另一条为 error 级别("response_code":504),最后一条为 warn 级别日志("response_code":404)。

>{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":204,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-04-30T08:47:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}
>
>
>
>
>{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":504,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-04-30T08:47:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}
>
>
>
>{"protocol":"HTTP/1.1","upstream_local_address":"172.20.32.97:33878","response_flags":"-","istio_policy_status":null,"trace_id":"5532224c1013b9ad6da1efe88778dd64","authority":"server:1338","method":"PUT","response_code":404,"duration":83,"upstream_service_time":"83","user_agent":"Jakarta Commons-HttpClient/3.1","bytes_received":103,"downstream_local_address":"172.21.2.130:1338","start_time":"2024-04-30T08:47:11.230Z","upstream_transport_failure_reason":null,"requested_server_name":null,"bytes_sent":0,"route_name":"routes","x_forwarded_for":"10.0.0.69,10.23.0.31","upstream_cluster":"outbound|1338|svc.cluster.local","request_id":"80ac7d31-a598-4dc8-bb74-1850593f61e4","downstream_remote_address":"10.23.0.31:0","path":"/api/dimensions/items","upstream_host":"172.20.9.101:1338"}

通过 DataKit 采集到 Kafka 的三条业务日志

使用 Pipeline 对业务日志进行字段提取

下图 protocol、response_code 以及 time 都是使用 Pipeline 提取后的效果。

相关推荐
java1234_小锋6 小时前
讲讲RabbitMQ 性能优化
kafka
码农爱java14 小时前
Kafka 之消息并发消费
spring boot·微服务·kafka·mq·消息中间件·并发消费
Mephisto.java18 小时前
【大数据学习 | kafka高级部分】kafka的快速读写
大数据·redis·sql·kafka·flume
Mephisto.java19 小时前
【大数据学习 | kafka高级部分】文件清除原理
大数据·hadoop·zookeeper·spark·kafka·hbase·flume
0_1_bits1 天前
【系统设计】高效的分布式系统:使用 Spring Boot 和 Kafka 实现 Saga 模式
spring boot·后端·云原生·架构·kafka·linq
不想睡觉的橘子君1 天前
【MQ】RabbitMQ、RocketMQ、kafka特性对比
kafka·rabbitmq·rocketmq
码农爱java2 天前
Kafka 之顺序消息
spring boot·分布式·微服务·kafka·mq·消息中间件·顺序消息
孤蓬&听雨2 天前
Kafka自动生产消息软件(自动化测试Kafka)
分布式·kafka·自动化·测试·生产者
Mephisto.java2 天前
【大数据学习 | kafka】producer之拦截器,序列化器与分区器
大数据·学习·kafka
mpb2 天前
centos7 kafka高可用集群安装及测试
kafka