kudu写入压力大时报错

19/05/18 16:53:12 INFO AsyncKuduClient: Invalidating location fd52e4f930bc45458a8f29ed118785e3(server002:7050) for tablet 4259921cdcca4776b37771659a8cafb3: Service unavailable: Soft memory limit exceeded (at 106.05% of capacity). See https://kudu.apache.org/releases/1.7.0-cdh5.16.1/docs/troubleshooting.html

这是个INFO日志,只是在内存超过hard的80%时拒绝写入,可能请求导致部分写入超时;

参数调优(增大以下参数):

  • 1 --memory_limit_hard_bytes
  • 2 --maintenance_manager_num_threads (Generally, the recommended ratio of maintenance manager threads to data directories is 1:3.)
  • 3 --block_cache_capacity_mb (memory_limit_hard_bytes的30%到50%)

原文:

Kudu has a hard and soft memory limit. The hard memory limit is the maximum amount a Kudu process is allowed to use, and is controlled by the --memory_limit_hard_bytes flag. The soft memory limit is a percentage of the hard memory limit, controlled by the flag memory_limit_soft_percentage and with a default value of 80%, that determines the amount of memory a process may use before it will start rejecting some write operations.

If the logs or RPC traces contain messages like

Service unavailable: Soft memory limit exceeded (at 96.35% of capacity)

then Kudu is rejecting writes due to memory backpressure. This may result in write timeouts. There are several ways to relieve the memory pressure on Kudu:

  • If the host has more memory available for Kudu, increase --memory_limit_hard_bytes.
  • Increase the rate at which Kudu can flush writes from memory to disk by increasing the number of disks or increasing the number of maintenance manager threads --maintenance_manager_num_threads. Generally, the recommended ratio of maintenance manager threads to data directories is 1:3.
  • Reduce the volume of writes flowing to Kudu on the application side.

Finally, on versions of Kudu prior to 1.8, check the value of --block_cache_capacity_mb. This setting determines the maximum size of Kudu’s block cache. While a higher value can help with read and write performance, do not raise --block_cache_capacity_mb above the memory pressure threshold, which is --memory_pressure_percentage (default 60%) of --memory_limit_hard_bytes, as this will cause Kudu to flush aggressively even if write throughput is low. Keeping --block_cache_capacity_mb below 50% of the memory pressure threshold is recommended. With the defaults, this means --block_cache_capacity_mb should not exceed 30% of --memory_limit_hard_bytes. On Kudu 1.8 and higher, servers will refuse to start if the block cache capacity exceeds the memory pressure threshold.

调优:

https://kudu.apache.org/docs/scaling_guide.html#memory

https://kudu.apache.org/docs/scaling_guide.html#_verifying_if_a_memory_limit_is_sufficient

参考:
https://kudu.apache.org/docs/troubleshooting.html
https://kudu.apache.org/docs/troubleshooting.html#memory_limits

【原创】大叔经验分享(58)kudu写入压力大时报错的更多相关文章

  1. 【原创】大叔经验分享(13)spark运行报错WARN Utils: Service 'sparkDriver' could not bind on port 0. Attempting port 1.

    本地运行spark报错 18/12/18 12:56:55 WARN Utils: Service 'sparkDriver' could not bind on port 0. Attempting ...

  2. 【原创】大叔经验分享(78)hive查询报错NoViableAltException

    Hive或spark中执行sql字符常量包含;时会报错,比如 select instr('abc;abc', ';'); 报错 NoViableAltException(-1@[147:1: sele ...

  3. 【原创】经验分享:一个小小emoji尽然牵扯出来这么多东西?

    前言 之前也分享过很多工作中踩坑的经验: 一个线上问题的思考:Eureka注册中心集群如何实现客户端请求负载及故障转移? [原创]经验分享:一个Content-Length引发的血案(almost.. ...

  4. 【原创】大叔经验分享(54)flume kudu sink运行一段时间kudu client报错

    flume kudu sink运行一段时间报错: 19/05/05 10:15:56 WARN client.ConnectToCluster: Error receiving a response ...

  5. 【原创】大叔经验分享(63)kudu vs parquet

    一 对比 存储空间对比: 查询性能对比: 二 设计方案 将数据拆分为:历史数据(hdfs+parquet+snappy)+ 近期数据(kudu),可以兼具各种优点: 1)整体低于10%的磁盘占用: 2 ...

  6. 【原创】大叔经验分享(53)kudu报错unable to find SASL plugin: PLAIN

    kudu安装后运行不正常,master中找不到任何tserver,查看tserver日志发现有很多报错: Failed to heartbeat to master:7051: Invalid arg ...

  7. 【原创】大叔经验分享(62)kudu副本数量

    kudu的副本数量是在表上设置,可以通过命令查看 # sudo -u kudu kudu cluster ksck $master ... Summary by table Name | RF | S ...

  8. 【原创】大叔经验分享(61)kudu rebalance报错

    kudu rebalance命令报错 terminate called after throwing an instance of 'std::regex_error' what(): regex_e ...

  9. 【原创】大叔经验分享(60)hive和spark读取kudu表

    从impala中创建kudu表之后,如果想从hive或spark sql直接读取,会报错: Caused by: java.lang.ClassNotFoundException: com.cloud ...

随机推荐

  1. 数据库 | SQL 诊断优化套路包,套路用的对,速度升百倍

    本文出自头条号老王谈运维,转载请说明出处. 前言 在DBA的日常工作中,调整个别性能较差的SQL语句是一项富有挑战性的工作.面对慢SQL,一些DBA会心烦,会沮丧,会束手无措,也会沉着冷静.斗智斗勇! ...

  2. P3951 小凯的疑惑

    P3951 小凯的疑惑 题解 题意也就是求解不能用 ax+by 表示的最大数 ans(a,b,x,y,都是正整数) 给定 a ( =7 ) ,  b ( =3 ) 我们可以把数轴非负半轴上的数按照a的 ...

  3. jQuery常用AJAX-API

    目的:简化客户端与服务端进行局部刷新的异步通讯 (1)取得服务端当前时间 简单形式:jQuery对象.load(url) 返回结果自动添加到jQuery对象代表的标签中间 如果是Servlet的话,采 ...

  4. oc 执行shell 脚本

    -(id) InvokingShellScriptAtPath :(NSString*) shellScriptPath { NSTask *shellTask = [[NSTask alloc]in ...

  5. Docker监控:最佳实践以及cAdvisor和Prometheus监控工具的对比

    在DockerCon EU 2015上,Brian Christner阐述了“Docker监控”的概况,分享了这方面的最佳实践和Docker stats API的指南,并对比了三个流行的监控方案:cA ...

  6. JAVA 基础编程练习题9 【程序 9 求完数】

    9 [程序 9 求完数] 题目:一个数如果恰好等于它的因子之和,这个数就称为"完数".例如 6=1+2+3.编程找出 1000 以内的 所有完数. package cskaoyan ...

  7. Java中的字节,字符与编码,解码

    ASCII编码 ASCII码主要是为了表示英文字符而设计的,ASCII码一共规定了128个字符的编码(0x00-0x7F),只占用了一个字节的后面7位,最前面的1位统一规定为0. ISO-8859-1 ...

  8. Android 单元测试学习计划

    网上查了一下Android单元测试相关的知识点,总结了一个学习步骤: 1. 什么是单元测试2. 单元测试正反面: 2.1. 重要性 2.2. 缺陷 2.3. 策略3. 单元测试的基础知识: 3.1. ...

  9. c/c++编码规范(1)--头文件

    最近工作稍微轻松一点,就再学习了一下编码规范.遂记些笔记,以便查阅. 这次学习的是Google 开源项目风格指南中文版,地址是:http://zh-google-styleguide.readthed ...

  10. 使用robotframework做接口测试5——一个用例中调多个接口

    凡是涉及一点点有接口关联的,都可能下一个接口需要上一个接口的某个返回值作为入参,最直接的例子,就是登录依赖.用接口做业务性的测试,也绝对离不开接口依赖的,业务都是一系列接口串联的结果,有时候一个接口操 ...