为了节约内存,使用IBM J9 JDK跑Hadoop 2.10.2,出现以下错误
org.apache.hadoop.security.KerberosAuthException: failure to login: javax.security.auth.login.LoginException: unable to find LoginModule class: com.ibm.security.auth.module.LinuxLoginModule
除了改源码,还有一个比较简单的处理方式,就是打补丁。使用:
-javaagent:<path_to_setvendor8.zip>
配置etc/hadoop/hadoop-env.sh
export HADOOP_OPTS="$HADOOP_OPTS -Djava.net.preferIPv4Stack=true"
#适配OPENJ9
export HADOOP_OPTS="$HADOOP_OPTS -javaagent:/opt/jdk-openj9-8/setvendor8.zip"
配置etc/hadoop/yarn-env.sh
YARN_OPTS="$YARN_OPTS -Dhadoop.root.logger=${YARN_ROOT_LOGGER:-INFO,console}"
YARN_OPTS="$YARN_OPTS -Dyarn.root.logger=${YARN_ROOT_LOGGER:-INFO,console}"
#openj9适配
YARN_OPTS="$YARN_OPTS -javaagent:/opt/jdk-openj9-8/setvendor8.zip"
然后start-all.sh 跑起来跑起来,内存占用相当舒适
setvendor8.zip下载地址 https://download.csdn.net/download/applebomb/88639890
MapReduce的错误
按照以上的方式,可以跑起来了,但是执行操作的时候还是会报错。MR操作是子进程进行的,所以一直找子进程怎样实现。经过数小时折腾,终于找到了办法,不确定是哪个OPTS的作用,索性把所有的OPTS都加上了,最后mapred-site.xml长这样:
<configuration>
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<property>
<name>mapred.child.java.opts</name>
<value>-javaagent:/opt/jdk-openj9-8/setvendor8.zip</value>
</property>
<property>
<name>yarn.app.mapreduce.am.command-opts</name>
<value>-javaagent:/opt/jdk-openj9-8/setvendor8.zip</value>
</property>
<property>
<name>yarn.app.mapreduce.am.admin-command-opts</name>
<value>-javaagent:/opt/jdk-openj9-8/setvendor8.zip</value>
</property>
</configuration>
然后可以配置hive,hive没有什么特别的,直接就可以跑起来
【2023-12-21】新增
如果要配置flink的继续看:
flink的配置主要是修改config.sh来适配,关键的改动点头部添加export环境导出javaagent:
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
################################################################################
export HADOOP_CLASSPATH=`hadoop classpath`
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export JVM_ARGS="$JVM_ARGS -javaagent:/opt/jdk-openj9-8/setvendor8.zip"
constructFlinkClassPath() {
local FLINK_DIST
local FLINK_CLASSPATH
这样在执行sql-client.sh embedded shell的时候就不会有报错了