来自：http://caiguangguang.blog.51cto.com/1652935/1384187

flume bucketpath的bug一例

测试的配置文件：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
agent-server1.sources= testtail
agent-server1.sinks = hdfs-sink
agent-server1.channels= hdfs-channel
agent-server1.sources.testtail.type = netcat
agent-server1.sources.testtail.bind = localhost
agent-server1.sources.testtail.port = 9999
agent-server1.sinks.hdfs-sink.hdfs.kerberosPrincipal = hdfs/_HOST@KERBEROS_HADOOP
agent-server1.sinks.hdfs-sink.hdfs.kerberosKeytab = /home/vipshop/conf/hdfs.keytab
agent-server1.channels.hdfs-channel.type = memory
agent-server1.channels.hdfs-channel.capacity = 200000000
agent-server1.channels.hdfs-channel.transactionCapacity = 10000
agent-server1.sinks.hdfs-sink.type = hdfs
agent-server1.sinks.hdfs-sink.hdfs.path = hdfs://bipcluster/tmp/flume/%Y%m%d
agent-server1.sinks.hdfs-sink.hdfs.rollInterval = 60
agent-server1.sinks.hdfs-sink.hdfs.rollSize = 0
agent-server1.sinks.hdfs-sink.hdfs.rollCount = 0
agent-server1.sinks.hdfs-sink.hdfs.threadsPoolSize = 10
agent-server1.sinks.hdfs-sink.hdfs.round = false
agent-server1.sinks.hdfs-sink.hdfs.roundValue = 30
agent-server1.sinks.hdfs-sink.hdfs.roundUnit = minute
agent-server1.sinks.hdfs-sink.hdfs.batchSize = 100
agent-server1.sinks.hdfs-sink.hdfs.fileType = DataStream
agent-server1.sinks.hdfs-sink.hdfs.writeFormat = Text
agent-server1.sinks.hdfs-sink.hdfs.callTimeout = 60000
agent-server1.sinks.hdfs-sink.hdfs.idleTimeout = 100
agent-server1.sinks.hdfs-sink.hdfs.filePrefix = ip
agent-server1.sinks.hdfs-sink.channel = hdfs-channel
agent-server1.sources.testtail.channels = hdfs-channel

在启动服务后，使用telnet进行测试，发现如下报错：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
14/03/24 18:03:07 ERROR hdfs.HDFSEventSink: process failed
java.lang.RuntimeException: Flume wasn't able to parse timestamp header in the event to resolve time based bucketing.
 Please check that you're correctly populating timestamp header (for example using TimestampInterceptor source interceptor).
        at org.apache.flume.formatter.output.BucketPath.replaceShorthand(BucketPath.java:160)
        at org.apache.flume.formatter.output.BucketPath.escapeString(BucketPath.java:343)
        at org.apache.flume.sink.hdfs.HDFSEventSink.process(HDFSEventSink.java:392)
        at org.apache.flume.sink.DefaultSinkProcessor.process(DefaultSinkProcessor.java:68)
        at org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:147)
        at java.lang.Thread.run(Thread.java:662)
Caused by: java.lang.NumberFormatException: null
        at java.lang.Long.parseLong(Long.java:375)
        at java.lang.Long.valueOf(Long.java:525)
        at org.apache.flume.formatter.output.BucketPath.replaceShorthand(BucketPath.java:158)
        ... 5 more
14/03/24 18:03:07 ERROR flume.SinkRunner: Unable to deliver event. Exception follows.
org.apache.flume.EventDeliveryException: java.lang.RuntimeException: Flume wasn't able to parse timestamp header in the event to
resolve time based bucketing. Please check that you're correctly populating timestamp header (for example using TimestampInterceptor source interceptor).
        at org.apache.flume.sink.hdfs.HDFSEventSink.process(HDFSEventSink.java:461)
        at org.apache.flume.sink.DefaultSinkProcessor.process(DefaultSinkProcessor.java:68)
        at org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:147)
        at java.lang.Thread.run(Thread.java:662)
Caused by: java.lang.RuntimeException: Flume wasn't able to parse timestamp header in the event to resolve time based bucketing. Please check that you're correctly populating timestamp header (for example using TimestampInterceptor source interceptor).
        at org.apache.flume.formatter.output.BucketPath.replaceShorthand(BucketPath.java:160)
        at org.apache.flume.formatter.output.BucketPath.escapeString(BucketPath.java:343)
        at org.apache.flume.sink.hdfs.HDFSEventSink.process(HDFSEventSink.java:392)
        ... 3 more
Caused by: java.lang.NumberFormatException: null
        at java.lang.Long.parseLong(Long.java:375)
        at java.lang.Long.valueOf(Long.java:525)
        at org.apache.flume.formatter.output.BucketPath.replaceShorthand(BucketPath.java:158)
        ... 5 more

从调用栈的信息来看，错误出在org.apache.flume.formatter.output.BucketPath类的replaceShorthand方法。
在org.apache.flume.sink.hdfs.HDFSEventSink类中，使用process方法来生成hdfs的url,其中主要是调用了BucketPath类的escapeString方法来进行字符的转换，并最终调用了replaceShorthand方法。
其中replaceShorthand方法的相关代码如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
  public static String replaceShorthand(char c, Map<String, String> headers,
      TimeZone timeZone, boolean needRounding, int unit, int roundDown) {
    String timestampHeader = headers.get("timestamp");
    long ts;
    try {
      ts = Long.valueOf(timestampHeader);
    } catch (NumberFormatException e) {
      throw new RuntimeException("Flume wasn't able to parse timestamp header"
        + " in the event to resolve time based bucketing. Please check that"
        + " you're correctly populating timestamp header (for example using"
        + " TimestampInterceptor source interceptor).", e);
    }
    if(needRounding){
      ts = roundDown(roundDown, unit, ts);
    }
........

从代码中可以看到，timestampHeader 的值如果取不到，在向ts赋值时就会报错。。
这其实是flume的一个bug，bug id：
https://issues.apache.org/jira/browse/FLUME-1419
解决方法有3个：
1.更改配置，更新hdfs文件的路径格式

1
agent-server1.sinks.hdfs-sink.hdfs.path = hdfs://bipcluster/tmp/flume

但是这样就不能按天来存放日志了
2.通过更改相关的代码
(patch:https://issues.apache.org/jira/secure/attachment/12538891/FLUME-1419.patch)
如果在headers中获取不到timestamp的值，就给它一个当前timestamp的值。
相关代码:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
     String timestampHeader = headers.get("timestamp");
     long ts;
     try {
      if (timestampHeader == null) {
        ts = System.currentTimeMillis();
      } else {
        ts = Long.valueOf(timestampHeader);
      }
     } catch (NumberFormatException e) {
       throw new RuntimeException("Flume wasn't able to parse timestamp header"
         + " in the event to resolve time based bucketing. Please check that"
         + " you're correctly populating timestamp header (for example using"
                  + " TimestampInterceptor source interceptor).", e);
}

3.为source定义基于timestamp的interceptors 
在配置中增加两行即可：

1
2
agent-server1.sources.testtail.interceptors = i1
agent-server1.sources.testtail.interceptors.i1.type = org.apache.flume.interceptor.TimestampInterceptor$Builder

一个技巧：
在debug flume的问题时，可以在flume的启动参数中设置把debug日志打到console中。

1
-Dflume.root.logger=DEBUG,console,LOGFILE

Flume wasn't able to parse timestamp header的更多相关文章

  1. 关于flume中涉及到时间戳的错误解决,Expected timestamp in the Flume even

    在搭建flume集群收集日志写入hdfs时发生了下面的错误: java.lang.NullPointerException: Expected timestamp in the Flume event ...

  2. 当我new class的时候，提示以下错误： Unable to parse template "Class" Error message: This template did not produce a Java class or an interface Error parsing file template: Unable to find resource 'Package Header.j

    你肯定修改过class的template模板,改回去就好了 #if (${PACKAGE_NAME} && ${PACKAGE_NAME} != "")packag ...

  3. Flume架构

    Flume是Cloudera提供的一个高可用的,高可靠的,分布式的海量日志采集.聚合和传输的系统: Flume 介绍 Flume是由cloudera软件公司产出的高可用.高可靠.分布式的海量日志收集系 ...

  4. netty 网关 flume 提交数据 去除透明 批处理 批提交 cat head tail 结合 管道显示行号

    D:\javaNettyAction\NettyA\src\main\java\com\test\HexDumpProxy.java package com.test; import io.netty ...

  5. TimeStamp

    private void Form1_Load(object sender, EventArgs e) { textBox1.Text= GenerateTimeStamp(System.DateTi ...

  6. Flume官方文档翻译——Flume 1.7.0 User Guide （unreleased version）（一）

    Flume 1.7.0 User Guide Introduction(简介) Overview(综述) System Requirements(系统需求) Architecture(架构) Data ...

  7. c# datetime与 timeStamp（unix时间戳） 互相转换

    /// <summary> /// Unix时间戳转为C#格式时间 /// </summary> /// <param name="timeStamp" ...

  8. webMagic解析淘宝cookie 提示Invalid cookie header

    webMagic解析淘宝cookie 提示Invalid cookie header 在使用webMagic框架做爬虫爬取淘宝极又家页面时候一直提醒cookie设置不可用如下图 淘宝的验证特别严重,c ...

  9. c# datetime与 timeStamp时间戳 互相转换

    将时间格式转化为一个int类型 // ::26时间转完后为:1389675686数字 为什么使用时间戳? 关于Unix时间戳,大概是这个意思,从1970年0时0分0秒开始到现在的秒数.使用它来获得的是 ...

随机推荐

  1. PostgreSQL SystemTap on Linux 转

    PostgreSQL 支持动态跟踪, 可以通过dtrace或者systemtap工具统计相关探针的信息. 安装systemtap yum install systemtap kernel-debugi ...

  2. My安装EclipseJS代码提示（Spket插件）

    最近须要大量使用JS来开发,可是MyEclipse2014自带的JS编辑器没有代码提示的功能,开发效率有点低,所以安装了一个Spket的插件,过程很easy,SVN插件的安装比这个更简单. Spket ...

  3. 【springboot+easypoi】一行代码搞定excel导入导出

    原文:https://www.jianshu.com/p/5d67fb720ece 开发中经常会遇到excel的处理,导入导出解析等等,java中比较流行的用poi,但是每次都要写大段工具类来搞定这事 ...

  4. [Unity Editor]10行代码搞定Hierarchy排序

    在日常的工作和研究中,当给我们的场景摆放过多的物件的时候,Hierarchy面板就会变得杂乱不堪.比如这样:    过多的层次结构充斥在里面,根层的物件毫无序列可言,整个层次面板显示非常的杂乱不堪,如 ...

  5. arcgis python 创建 SQLite 数据库

    # -*- ################# """ Tool name: Create SQLite Database Source: CreateSQLiteDat ...

  6. 【微信小程序】在微信开发工具上七牛云的图片可以看到，但是在真机上看不到的原因解决

    在开发微信小程序过程中,在微信开发者工具上,七牛云的图片都可以展示出来,但是在真机上,七牛云的图片却展示不出来,也没有报404找不到或者不能加载图片的问题, 必须保证: 1.图片是用image加载的: ...

  7. Java中List效率的比较

    Java Collections Framework(JCF) 是Java SE中一个基本的类集,几乎所有的项目都会用到,其中的List 则是JCF中最最常用的一个接口.围绕List 接口,有很多实现 ...

  8. Log4j输出格式控制

    参数说明例子 %c 列出logger名字空间的全称,如果加上{<层数>}表示列出从最内层算起的指定层数的名字空间 log4j配置文件参数举例 输出显示媒介 假设当前logger名字空间是& ...

  9. 0, \0, NULL

    字符串.字符数组输入.输出与'\0'的问题 原创首发,欢迎转载! 作者按 字符串.字符数组以"%s"格式输入时,以遇到'空格'为这个字符串输入结束. 字符串.字符数组以" ...

  10. poj 1469 COURSES 题解

    COURSES Time Limit: 1000MS   Memory Limit: 10000K Total Submissions: 21515   Accepted: 8455 Descript ...