java - 配置单元 : Tez how to increase AM Container memory

Question

我正在尝试运行配置单元查询

set yarn.nodemanager.vmem-check-enabled=false;
set hive.strict.checks.cartesian.product=false;
select count(*)
from db1.tb1 a
where a.col1='2015-07-13'
and a.col2='val'
and exists (
         select 1 from db2.tb2 b
         where b.col1='2015-07-13'
           and b.col2='val'
           and size(a.col3)=size(b.col3));

但我不断得到

Application application_1585768406438_10833 failed 2 times due to AM Container for appattempt_1585768406438_10833_000002 exited with  exitCode: -104
Failing this attempt.Diagnostics: [2020-04-26 19:47:06.497]Container [pid=30016,containerID=container_e03_1585768406438_10833_02_000001] is running beyond physical memory limits. Current usage: 1.0 GB of 1 GB physical memory used; 2.8 GB of 2.1 GB virtual memory used. Killing container.
Dump of the process-tree for container_e03_1585768406438_10833_02_000001 :s
--------|- PID PPID PGRPID SESSID CMD_NAME USER_MODE_TIME(MILLIS) SYSTEM_TIME(MILLIS) VMEM_USAGE(BYTES) RSSMEM_USAGE(PAGES) FULL_CMD_LINE
    |- 30101 30016 30016 30016 (java) 11364 769 3041062912 264859 /usr/lib/jvm/java-8-openjdk-amd64/bin/java -Xmx819m -Djava.io.tmpdir=/hadoop/yarn/nm-local-dir/usercache/a0t00wf/appcache/application_1585768406438_10833/container_e03_1585768406438_10833_02_000001/tmp -server -Djava.net.preferIPv4Stack=true -Dhadoop.metrics.log.level=WARN -XX:+PrintGCDetails -verbose:gc -XX:+PrintGCTimeStamps -XX:+UseNUMA -XX:+UseParallelGC -Dlog4j.configuratorClass=org.apache.tez.common.TezLog4jConfigurator -Dlog4j.configuration=tez-container-log4j.properties -Dyarn.app.container.log.dir=/var/log/hadoop-yarn/userlogs/application_1585768406438_10833/container_e03_1585768406438_10833_02_000001 -Dtez.root.logger=INFO,CLA -Dsun.nio.ch.bugLevel= org.apache.tez.dag.app.DAGAppMaster --session

我尝试将其作为mapreduce而不是tez运行，但我面临同样的问题。我尝试了以下设置

泰兹

set tez.am.resource.memory.mb=4096
set tez.task.resource.memory.mb=4096
set tez.am.java.opts=-Xmx6144m;
set tez.am.resource.memory.mb=4096;
set hive.tez.container.size=4096;

先生

set hive.execution.engine=mr;
set mapreduce.map.memory.mb=4096;
set mapreduce.reduce.memory.mb=4096;
set mapreduce.map.java.opts=-Xmx6144m;
set mapreduce.reduce.java.opts=-Xmx6144m;

还有这些 Yarn 设置

set yarn.nodemanager.vmem-check-enabled=false;
set yarn.nodemanager.resource.memory-mb=98304;
set yarn.scheduler.minimum-allocation-mb=8192;
set yarn.scheduler.maximum-allocation-mb=98304;
set yarn.nodemanager.vmem-pmem-ratio=9;

但我总是遇到同样的错误。如何增加 AM 容器的内存限制？我怀疑问题在于 java 内存的一些限制，如堆栈跟踪中所示

/usr/lib/jvm/java-8-openjdk-amd64/bin/java -Xmx819m

是这样吗？如果是这样，我应该如何增加 tez/hive 使用的 java 内存？

Answer 1

首先您需要了解堆是容器的子集。您的堆内存应约为容器内存的 80%。

set hive.execution.engine=mr;  
set mapreduce.map.memory.mb=4096;  -- this is container memory
set mapreduce.reduce.memory.mb=4096;

以下值是错误的。它们必须小于 4096，否则容器运行时总会出现超出内存限制的问题。

set mapreduce.map.java.opts=-Xmx6144m;  -- this is heap memory
set mapreduce.reduce.java.opts=-Xmx6144m;

而是将它们设置为:

set mapreduce.map.java.opts=-Xmx3276m;   -- (80% of 4096)
set mapreduce.reduce.java.opts=-Xmx3276m;

这是一篇关于理解这些术语的好文章:https://community.cloudera.com/t5/Community-Articles/Demystify-Apache-Tez-Memory-Tuning-Step-by-Step/ta-p/245279