ELK+Filebeat+Kafka,SpringBoot异常日志收集

概述

FileBeat + Kafka + ELK，收集SpringBoot项目的日志输出，有几个地方需要注意：

• 项目日志输出格式化
• Filebeat将多行日志汇集成一行
• Logstash将项目日志格式化为json

SpringBoot项目日志输出格式化

logback-spring.xml

    
        
            
        
        
            
        
    
    
        
            [%d] [%p] [%t] [%c] - %m%n
            UTF-8
        
    
...
    
        
    

xml配置文件中，[%d] [%p] [%t] [%c] - %m%n 确定了日志的规则，实际的日志输出格式如下：

[2021-04-25 09:50:58,255] [INFO] [RMI TCP Connection(23)-10.10.20.248] [org.springframework.web.servlet.DispatcherServlet] - Initializing Servlet 'dispatcherServlet'
[2021-04-25 09:50:58,324] [INFO] [RMI TCP Connection(23)-10.10.20.248] [org.springframework.web.servlet.DispatcherServlet] - Completed initialization in 69 ms
[2021-04-25 09:50:58,565] [WARN] [RMI TCP Connection(24)-10.10.20.248] [org.springframework.boot.actuate.elasticsearch.ElasticsearchHealthIndicator] - Elasticsearch health check failed
org.elasticsearch.client.transport.NoNodeAvailableException: None of the configured nodes are available: [{
     #transport#-1}{
     haxOm4fCQzaVW60mmuR-4w}{
     10.10.50.84}{
     10.10.50.84:9300}]
	at org.elasticsearch.client.transport.TransportClientNodesService.ensureNodesAreAvailable(TransportClientNodesService.java:349)
	at org.elasticsearch.client.transport.TransportClientNodesService.execute(TransportClientNodesService.java:247)
	at org.elasticsearch.client.transport.TransportProxyClient.execute(TransportProxyClient.java:60)
	at org.elasticsearch.client.transport.TransportClient.doExecute(TransportClient.java:381)
	at org.elasticsearch.client.support.AbstractClient.execute(AbstractClient.java:407)
	at org.elasticsearch.client.support.AbstractClient.execute(AbstractClient.java:396)
	at org.elasticsearch.client.support.AbstractClient$ClusterAdmin.execute(AbstractClient.java:708)
	at org.elasticsearch.client.support.AbstractClient$ClusterAdmin.health(AbstractClient.java:730)
	at org.springframework.boot.actuate.elasticsearch.ElasticsearchHealthIndicator.doHealthCheck(ElasticsearchHealthIndicator.java:79)
	at org.springframework.boot.actuate.health.AbstractHealthIndicator.health(AbstractHealthIndicator.java:84)
	at org.springframework.boot.actuate.health.CompositeHealthIndicator.health(CompositeHealthIndicator.java:98)
	at org.springframework.boot.actuate.health.HealthEndpoint.health(HealthEndpoint.java:50)
	at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
	at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)

配置Filebeat汇集换行的日志为一行

vi filebeat.yml

#=========================== Filebeat inputs =============================
filebeat.inputs:
- type: log
  enabled: true
  # lms 8301
  paths:
    - /var/logs/xapi/8301/*.log
  tail_files: true
  fields:
    logtopic: elk-lms-8301
  ### Multiline options
  multiline.pattern: '^\['
  multiline.negate: true
  multiline.match: after

#================================ Kafka ========================================
output.kafka:
  enabled: true
  # initial brokers for reading cluster metadata
  hosts: ["10.10.20.26:9092"]

  # message topic selection + partitioning
  #topic: elk-lms-xapi
  topic: '%{[fields.logtopic]}'
  partition.round_robin:
    reachable_only: false

  required_acks: 1
  compression: gzip
  max_message_bytes: 1000000
name: 10.10.20.26_filebeat

这里的multiline.pattern，表示以‘[’开始的行为新一行。新一行之前的输出，算为当前行的内容。

Logstash格式化项目日志

vi logstash-8301.conf

# Sample Logstash configuration for creating a simple
# Beats -> Logstash -> Elasticsearch pipeline.
input {
     
    kafka {
     
        bootstrap_servers => "10.10.20.26:9092"
        topics => ["elk-lms-8301"]
#        consumer_threads => 5
#        decorate_events => true
        auto_offset_reset => "latest"
#        auto_offset_reset => "earliest"
        client_id => "client-xapi-8301"
        group_id => "group-xapi-8301"
    }


}

filter {
     
    # 将message转为json格式
        grok {
       
            match => {
     
                "message" => "\[%{TIMESTAMP_ISO8601:log_time}\]\s\[%{LOGLEVEL:log_level}\]\s\[%{GREEDYDATA:log_from}\]\s\[%{JAVACLASS:log_classname}\]\s\-\s%{GREEDYDATA:log_detail}"
            }
        }
        
        # 正则替换日志中的换行符, 最后一个参数是双引号包围的回车键
        mutate {
     
            gsub => ["log_detail", "\\n", "
            "] 
        }


        # 移动log_detail字符串中的系统属性 begin
        if [log_detail] =~ "\"beat\"" {
     
            grok {
     
                match => {
      
                    "log_detail" => "(?(.*)(?=\"beat\")/?)"
                }
                overwrite => ["log_detail"]
            }
        }       
        if [log_detail] =~ "\"offset\"" {
     
            grok {
     
                match => {
     
                    "log_detail" => "(?(.*)(?=\"offset\")/?)"
                }
                overwrite => ["log_detail"]
            }
        }       
        if [log_detail] =~ "\"source\"" {
     
            grok {
     
                match => {
     
                    "log_detail" => "(?(.*)(?=\"source\")/?)"
                }
                overwrite => ["log_detail"]
            }
        }       
        if [log_detail] =~ "\"fields\"" {
     
            grok {
     
                match => {
     
                    "log_detail" => "(?(.*)(?=\"fields\")/?)"
                }
                overwrite => ["log_detail"] 
            }
        }       
        if [log_detail] =~ "\"input\"" {
     
            grok {
     
                match => {
     
                    "log_detail" => "(?(.*)(?=\"input\")/?)"
                }
                overwrite => ["log_detail"]
            }
        }
        if [log_detail] =~ "\"prospector\"" {
     
            grok {
     
                match => {
     
                    "log_detail" => "(?(.*)(?=\"prospector\")/?)"
                }
                overwrite => ["log_detail"]
            }
        }
        # 移动log_detail字符串中的系统属性 end

}

output {
     
    # 处理后的日志入es
    if [fields][logtopic] == "elk-lms-8301" {
     
        elasticsearch {
     
            hosts => "10.10.20.26:9200"
            index => "elk-lms-8301"
        }
    }

    stdout {
     
        codec => rubydebug
    }
}

这里的关键配置是grok部分。
[%{TIMESTAMP_ISO8601:log_time}]\s[%{LOGLEVEL:log_level}]\s[%{GREEDYDATA:log_from}]\s[%{JAVACLASS:log_classname}]\s-\s%{GREEDYDATA:log_detail} 逐节解析如下：

• [%{TIMESTAMP_ISO8601:log_time}] ：表示 [时间戳]，如[2021-04-25 09:50:58,255]
• [%{LOGLEVEL:log_level}]：表示日志级别，如[INFO]
• [%{GREEDYDATA:log_from}]：表示来源，如[RMI TCP Connection(23)-10.10.20.248]
• [%{JAVACLASS:log_classname}]：表示java全路径的类名，如org.springframework.web.servlet.DispatcherServlet
• {GREEDYDATA:log_detail}：表示日志内容，如Initializing Servlet ‘dispatcherServlet’

所以，象下面的日志：

[2021-04-25 09:50:58,255] [INFO] [RMI TCP Connection(23)-10.10.20.248] [org.springframework.web.servlet.DispatcherServlet] - Initializing Servlet 'dispatcherServlet'

就会被格式化为：

{
     
  "classname": "org.springframework.web.servlet.DispatcherServlet",
  "level": "INFO",
  "from": "RMI TCP Connection(23)-10.10.20.248",
  "detail": "Initializing Servlet 'dispatcherServlet'",
  "timestamp": "2021-04-25 09:50:58,255"
}

在kibana的Grok Debugger中，可以看到格式化的效果：

另，按如下方式启动Logstash，可以在修改配置文件时，自动重启Logstash，比较适合调试

nohup /home/wender/app/logstash-6.4.3/bin/logstash -f /home/wender/app/logstash-6.4.3/config/logstash.conf --config.reload.automatic &

---

参考资料

Logstash Grok patterns
ELK集群+Filebeat+Kafka
Kibana 用户手册
开始使用 Elasticsearch