k8s容器环境收集应用日志到已有的ELK日志平台

Tags： k8s环境下的容器日志收集
K8S环境下面如何收集应用日志
===
在本文中重点讲一下K8S容器环境中如何收集容器的日志；

1. 容器日志收集方案的选择：

  在K8S集群中，容器的日志收集方案一般有三种；第一种方案是通过在每一个k8s节点安装日志收集客户端软件，比如fluentd。这种方案不好的一点是应用的日志必须输出到标准输出，并且是通过在每一台计算节点的/var/log/containers目录下面的日志文件，这个日志文件的名称是这种格式user-center-765885677f-j68zt_default_user-center-0867b9c2f8ede64cebeb359dd08a6b05f690d50427aa89f7498597db8944cccc.log，文件名称有很多随机字符串，很难和容器里面的应用对应起来。并且在网上看到别人说这个里面的日志，对于JAVA的报错内容没有多行合并，不过我还没有测试过此方案。

  第二种方案就是在应用的pods里面在运行一个sidecar container（边角容器），这个容器会和应用的容器挂载同一个volume日志卷。比如这个sidecar容器可以是filebeat或者flunetd等；这种方案不足之处是部署了sidecar , 所以会消耗资源 , 每个pod都要起一个日志收集容器。
  第三种方案就是直接将应用的日志收集到kafka，然后通过kafka再发送到logstash,再处理成json格式的日志发送到es集群，最后在kibana展示。我实验的就是这种方案。通过修改logsbak配置文件实现了日志直接发送到kafka缓存的功能；下面直接看配置了

1. logsbak配置：

<?xml version="1.0" encoding="UTF-8"?>
<configuration>
    <jmxConfigurator/>  <!-- 动态加载-->

    <property name="log-path" value="/apptestlogs" />  <!-- 统一 /applogs 下面 -->
    <property name="app-name" value="test" />  <!-- 应用系统名称 -->
    <property name="filename" value="test-test" />  <!---日志文件名，默认组件名称 -->
    <property name="dev-group-name" value="test" /> <!-- 开发团队名称 -->

    <conversionRule conversionWord="traceId"  converterClass="org.lsqt.components.log.logback.TraceIdConvert"/>

    <!-- 根据实际情况修改变量 end-->
    -<appender name="consoleAppender" class="ch.qos.logback.core.ConsoleAppender">
    <!-- 典型的日志pattern -->
        <!-- -<encoder>-->
          <!--<pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%traceId] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>-->
        <!--</encoder>-->
    -<encoder class="ch.qos.logback.core.encoder.LayoutWrappingEncoder">
    <layout class="org.apache.skywalking.apm.toolkit.log.logback.v1.x.TraceIdPatternLogbackLayout">
        <pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%tid] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>
    </layout>
    </encoder>
    </appender>

    -<appender name="fileAppender" class="ch.qos.logback.core.rolling.RollingFileAppender">
    <file>${log-path}/${app-name}/${filename}.log</file>
    -<rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">
    <fileNamePattern>/${log-path}/${app-name}/${filename}.%d{yyyy-MM-dd}.%i.log</fileNamePattern>
    <maxHistory>15</maxHistory>
    <!--用来指定日志文件的上限大小，例如设置为300M的话，那么到了这个值，就会删除旧的日志。-->
    <timeBasedFileNamingAndTriggeringPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedFNATP">
        <maxFileSize>300MB</maxFileSize>
    </timeBasedFileNamingAndTriggeringPolicy>
</rollingPolicy>
    <!-- -<encoder>-->
    <!--<pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%traceId] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>-->
<!--</encoder>-->
    -<encoder class="ch.qos.logback.core.encoder.LayoutWrappingEncoder">
    <layout class="org.apache.skywalking.apm.toolkit.log.logback.v1.x.TraceIdPatternLogbackLayout">
        <pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%tid] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>
    </layout>
</encoder>
</appender>
    <appender name="errorAppender" class="ch.qos.logback.core.rolling.RollingFileAppender">
        <file>${log-path}/${app-name}/${filename}-error.log</file>
        <rollingPolicy class="ch.qos.logback.core.rolling.TimeBasedRollingPolicy">
            <fileNamePattern>/${log-path}/${app-name}/${filename}-error.%d{yyyy-MM-dd}.%i.log</fileNamePattern>
            <timeBasedFileNamingAndTriggeringPolicy class="ch.qos.logback.core.rolling.SizeAndTimeBasedFNATP">
                <maxFileSize>300MB</maxFileSize>
            </timeBasedFileNamingAndTriggeringPolicy>
            <maxHistory>15</maxHistory>
        </rollingPolicy>
        <!--<encoder class="ch.qos.logback.classic.encoder.PatternLayoutEncoder">-->
            <!--<pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%traceId] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>-->
        <!--</encoder>-->
        <encoder class="ch.qos.logback.core.encoder.LayoutWrappingEncoder">
        <layout class="org.apache.skywalking.apm.toolkit.log.logback.v1.x.TraceIdPatternLogbackLayout">
            <pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%tid] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>
        </layout>
    </encoder>
        <filter class="ch.qos.logback.classic.filter.LevelFilter">
            <level>ERROR</level>
            <onMatch>ACCEPT</onMatch>
            <onMismatch>DENY</onMismatch>
        </filter>
    </appender>

    <!-- This example configuration is probably most unreliable under
    failure conditions but wont block your application at all -->
    <appender name="very-relaxed-and-fast-kafka-appender" class="com.github.danielwegener.logback.kafka.KafkaAppender">
        <encoder class="ch.qos.logback.classic.encoder.PatternLayoutEncoder">
           <pattern>[%date{ISO8601}] [%level] %logger{80} [%thread] [%tid] ${dev-group-name} ${app-name} Line:%-3L - %msg%n</pattern>
        </encoder>
        <topic>elk-stand-sit-fkp-eureka</topic>
        <!-- we don't care how the log messages will be partitioned  -->
        <keyingStrategy class="com.github.danielwegener.logback.kafka.keying.NoKeyKeyingStrategy" />

        <!-- use async delivery. the application threads are not blocked by logging -->
        <deliveryStrategy class="com.github.danielwegener.logback.kafka.delivery.AsynchronousDeliveryStrategy" />

        <!-- each <producerConfig> translates to regular kafka-client config (format: key=value) -->
        <!-- producer configs are documented here: https://kafka.apache.org/documentation.html#newproducerconfigs -->
        <!-- bootstrap.servers is the only mandatory producerConfig -->
        <producerConfig>bootstrap.servers=192.168.1.12:9092,192.168.1.14:9092,192.168.1.15:9092</producerConfig>
        <!-- don't wait for a broker to ack the reception of a batch.  -->
        <producerConfig>acks=0</producerConfig>
        <!-- wait up to 1000ms and collect log messages before sending them as a batch -->
        <producerConfig>linger.ms=1000</producerConfig>
        <!-- even if the producer buffer runs full, do not block the application but start to drop messages -->
        <producerConfig>max.block.ms=0</producerConfig>
        <!-- define a client-id that you use to identify yourself against the kafka broker -->
        <producerConfig>client.id=${HOSTNAME}-${CONTEXT_NAME}-logback-relaxed</producerConfig>
        <!-- define All log messages that cannot be delivered fast enough will then immediately go to the fallback appenders -->
        <producerConfig>block.on.buffer.full=false</producerConfig>

         <!-- this is the fallback appender if kafka is not available. -->
        <appender-ref ref="consoleAppender" />
    </appender>

    <root level="debug">
        <appender-ref ref="very-relaxed-and-fast-kafka-appender" /> 
        <appender-ref ref="fileAppender"/>
        <appender-ref ref="consoleAppender"/>
        <appender-ref ref="errorAppender"/>

    </root>
</configuration>

###2. 针对logsbak配置说明：###

1. logsbak直接发送日志到kafka有几种方式，一种是异步模式，一种是同步模式。异步模式的意思就是如果kafka因为网络情况出现故障，则阻塞发送日志或者直接将日志发送到后备存储，比如后备存储是发送到日志文件；同步模式的意思就是即使kafka出现网络情况不可达，则就会影响到日志线程，进而影响到应用的性能。不过这个模式的我没有测试过；配置如下：

    <!-- This example configuration is more restrictive and will try to ensure that every message
     is eventually delivered in an ordered fashion (as long the logging application stays alive) -->
    <appender name="very-restrictive-kafka-appender" class="com.github.danielwegener.logback.kafka.KafkaAppender">
        <encoder class="ch.qos.logback.classic.encoder.PatternLayoutEncoder">
            <pattern>%d{HH:mm:ss.SSS} [%thread] %-5level %logger{36} - %msg%n</pattern>
        </encoder>

        <topic>important-logs</topic>
        <!-- ensure that every message sent by the executing host is partitioned to the same partition strategy -->
        <keyingStrategy class="com.github.danielwegener.logback.kafka.keying.HostNameKeyingStrategy" />
        <!-- block the logging application thread if the kafka appender cannot keep up with sending the log messages -->
        <deliveryStrategy class="com.github.danielwegener.logback.kafka.delivery.BlockingDeliveryStrategy">
            <!-- wait indefinitely until the kafka producer was able to send the message -->
            <timeout>0</timeout>
        </deliveryStrategy>

        <!-- each <producerConfig> translates to regular kafka-client config (format: key=value) -->
        <!-- producer configs are documented here: https://kafka.apache.org/documentation.html#newproducerconfigs -->
        <!-- bootstrap.servers is the only mandatory producerConfig -->
        <producerConfig>bootstrap.servers=localhost:9092</producerConfig>
        <!-- restrict the size of the buffered batches to 8MB (default is 32MB) -->
        <producerConfig>buffer.memory=8388608</producerConfig>

        <!-- If the kafka broker is not online when we try to log, just block until it becomes available -->
        <producerConfig>metadata.fetch.timeout.ms=99999999999</producerConfig>
        <!-- define a client-id that you use to identify yourself against the kafka broker -->
        <producerConfig>client.id=${HOSTNAME}-${CONTEXT_NAME}-logback-restrictive</producerConfig>
        <!-- use gzip to compress each batch of log messages. valid values: none, gzip, snappy  -->
        <producerConfig>compression.type=gzip</producerConfig>

        <!-- Log every log message that could not be sent to kafka to STDERR -->
        <appender-ref ref="STDERR"/>
    </appender>  

通过配置logsbak直接输出到kafka，并且使用异步模式，就成功的在kibana里面看到了容器的日志了；

博文的更详细内容请关注我的个人微信公众号 “云时代IT运维”,本公众号旨在共享互联网运维新技术，新趋势； 包括IT运维行业的咨询，运维技术文档分享。重点关注devops、jenkins、zabbix监控、kubernetes、ELK、各种中间件的使用，比如redis、MQ等；shell和python等运维编程语言；本人从事IT运维相关的工作有十多年。2008年开始专职从事Linux/Unix系统运维工作；对运维相关技术有一定程度的理解。本公众号所有博文均是我的实际工作经验总结，基本都是原创博文。我很乐意将我积累的经验、心得、技术与大家分享交流！希望和大家在IT运维职业道路上一起成长和进步；