兩臺用的都是ubuntu
IP | 主機名 |
---|---|
192.168.22.137 | spark-master |
192.168.22.150 | spark-slave1 |
更改主機名
確定每個節點的主機名與它在集羣中所處的位置相同
如果不同,需要修改vi /etc/hostname
重啓生效
可能需要些安裝某些工具包
- 更換sources源
vi /etc/apt/sources.list
deb http://mirrors.aliyun.com/ubuntu/ trusty main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-security main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-updates main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb http://mirrors.aliyun.com/ubuntu/ trusty-backports main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-security main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-updates main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-proposed main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/ trusty-backports main restricted universe multiverse
apt install net-tools
apt-get install iputils-ping
修改各主機的hosts文件
vi /etc/hosts
添加以下內容
192.168.22.137 spark-master
192.168.22.150 spark-slave1
SSH免密登錄
我看了網上別人的說只需要安裝server,但是我沒有成功,我安裝了server和client才行
apt-get install openssh-client
apt-get install openssh-server
# 啓動ssh服務
etc/init.d/ssh start
*關於ssh服務可以參照這個鏈接
http://linux.it.net.cn/e/server/ssh/2015/0501/14838.html*
緊接着就是配置各主機的免密登錄
- 所有的主機都需要生成私鑰和公鑰(直接回車)
ssh-keygen -t rsa
- 將所有主機的
~/.ssh/id_rsa.pub
都要放在master節點的~/.ssh/
目錄下(最好更改用以區分)
我使用的lrzsz
工具(有點笨)。之後再主機執行
你也可以使用scp ~/.ssh/id_rsa.pub root@<hostname|ip>:~/.ssh/id_rsa.pub.slave1
- 將所有公鑰加到用於認證的公鑰文件authorized_keys中
cat ~/.ssh/id_rsa.pub* >> ~/.ssh/authorized_keys
此時~/.ssh/
再將~/.ssh/authorized_keys
拷貝其他節點,到此個主機就完成了免密登錄
驗證一下:
ssh spark-master
ssh spark-slave1
如果出現如下,就說明你成功了
所需環境配置
準備軟件
用的版本不是最新的,看個人需要,但要保證各軟件的版本要互相支持
.
├── hadoop-2.6.5.tar.gz
├── jdk-8u171-linux-x64.tar.gz
├── scala-2.10.4.tgz
└── spark-1.6.3-bin-hadoop2.6.tgz
可直接去各大官網下載,如果你想省事,直接從網盤裏下載也行
https://pan.baidu.com/s/1vSu-6OTMvkROCBsiJwbMWQ
統一配置環境
我將所有的軟件放在/spark/software/
目錄下、解壓與修改文件名,然後統一配置環境變量
vi /etv/profile
假如如下內容
export JAVA_HOME=/spark/software/java
export JRE_HOME=$JAVA_HOME/jre
export PATH=$JAVA_HOME/bin:$JAVA_HOME/jre/bin:$PATH
export CLASSPATH=$CLASSPATH:.:$JAVA_HOME/lib:$JAVA_HOME/jre/lib
export SCALA_HOME=/spark/software/scala
export PATH=$PATH:$SCALA_HOME/bin
export HADOOP_HOME=/spark/software/hadoop
export HADOOP_CONF_DIR=${HADOOP_HOME}/etc/hadoop
export YARN_HOME=/spark/software/hadoop
export YARN_CONF_DIR=${YARN_HOME}/etc/hadoop
然後執行source /etc/profile
另其生效
然後測試,出現如下說明成功了
當然你也可以吧hadoop和spark假如path中,這樣就可以隨時使用hdfs
和spark-submit
命令了
其他主機做同樣的操作
提示:各主機最好都統一路徑,這樣修改一個文件,然後將文件直接遠程拷貝到其他主機上就行了
HADOOP配置
在/spark/software/hadoop/etc/hadoop
目錄下需要配置以下幾個文件:
hadoop-env.sh,
yarn-env.sh,
slaves,
core-site.xml,
hdfs-site.xml,
maprd-site.xml,
yarn-site.xml
hadoop-env.sh
export JAVA_HOME=/spark/software/java
yarn-env.sh
export JAVA_HOME=/spark/software/java
slaves
spark-slave1
(這裏我只添加了一個slave,你也可以把master加上去)
core-site.xml
添加如下:
<property>
<name>fs.defaultFS</name>
<value>hdfs://spark-master:9000/</value>
</property>
<property>
<name>hadoop.tmp.dir</name>
<value>file:/spark/software/hadoop/tmp</value>
</property>
hdfs-site.xml
添加如下:
<property>
<name>dfs.namenode.secondary.http-address</name>
<value>spark-master:9001</value>
</property>
<property>
<name>dfs.namenode.name.dir</name>
<value>file:/spark/software/hadoop/dfs/name</value>
</property>
<property>
<name>dfs.datanode.data.dir</name>
<value>file:/spark/software/hadoop/dfs/data</value>
</property>
<property>
<name>dfs.replication</name>
<value>3</value>
</property>
maprd-site.xml
添加如下
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
<!-- 下面的視情況而配置你可以先只配置上面的即可 -->
<property>
<name>mapreduce.map.memory.mb</name>
<value>1536</value>
</property>
<property>
<name>mapreduce.map.java.opts</name>
<value>-Xmx1024M</value>
</property>
<property>
<name>mapreduce.reduce.memory.mb</name>
<value>1024</value>
</property>
<property>
<name>mapreduce.reduce.java.opts</name>
<value>-Xmx1024M</value>
</property>
yarn-site.xml
添加如下
<property>
<name>yarn.nodemanager.aux-services</name>
<value>mapreduce_shuffle</value>
</property>
<property>
<name>yarn.nodemanager.aux-services.mapreduce.shuffle.class</name>
<value>org.apache.hadoop.mapred.ShuffleHandler</value>
</property>
<property>
<name>yarn.resourcemanager.address</name>
<value>spark-master:8032</value>
</property>
<property>
<name>yarn.resourcemanager.scheduler.address</name>
<value>spark-master:8030</value>
</property>
<property>
<name>yarn.resourcemanager.resource-tracker.address</name>
<value>spark-master:8035</value>
</property>
<property>
<name>yarn.resourcemanager.admin.address</name>
<value>spark-master:8033</value>
</property>
<property>
<name>yarn.resourcemanager.webapp.address</name>
<value>spark-master:8088</value>
</property>
<!-- 下面是情況而定 具體可以參考這裏 http://blog.javachen.com/2015/06/05/yarn-memory-and-cpu-configuration.html-->
<property>
<name>yarn.nodemanager.resource.memory-mb</name>
<value>2000</value>
</property>
<property>
<name>yarn.scheduler.maximum-allocation-mb</name>
<value>2000</value>
</property>
以上配置完畢之後,要同步到其他主機上,因爲配置了免密,可以這樣操作
scp /spark/software/hadoop/etc/hadoop/ root@spark-slave1:/spark/software/hadoop/etc/hadoop/
HADOOP啓動
進入/spark/software/hadoop
目錄下
- 格式化namenode
bin/hfds namenode -format
當出現“successful”的字樣,就說明成功了
啓動dfs
sbin/start-dfs.sh
啓動yarn
sbin/start-yarn.sh
接下來驗證,spark-master 執行jps,有以下幾個進程
27570 SecondaryNameNode
27720 ResourceManager
27356 NameNode
32476 Jps
每個slave上應該有以下幾個進程
18324 DataNode
18489 NodeManager
21055 Jps
可以在任意一臺主機上的瀏覽器輸入
http://spark-master:8088/cluster/nodes yarn管理界面
http://spark-master:50070 hdfs頁面
spark 環境
在/spark/software/spark/conf
目錄下修改spark-env.sh
(需先拷貝spark-env.sh.template)文件
export SCALA_HOME=spark/software/scala
export JAVA_HOME=/spark/software/java
export HADOOP_HOME=/spark/software/hadoop
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
SPARK_MASTER_IP=spark-master
SPARK_LOCAL_DIRS=/spark/software/spark
SPARK_DRIVER_MEMORY=1G
同樣在slaves文件中添加子節點
spark-slave1
同樣將這兩個文件發送到其他主機對應位置
然後在/spark/software/spark
目錄下執行
sbin/start-all.sh
瀏覽器輸入
http://spark-master:8080/
到此就搭建好了
運行
spark提供了很多例子,我們直接運行即可
# 本地運行
bin/spark-submit examples/src/main/python/pi.py 10 --master local[4]
# Spark Standalone 集羣模式運行
bin/spark-submit examples/src/main/python/pi.py 10 --master spark://spark-master:7077
# Spark on YARN 集羣上 yarn-cluster 模式運行
bin/spark-submit \
--class com.spark.WordCount \
--master yarn-client \
--driver-memory 1g \
--executor-cores 1 \
simple/word-count-1.0-SNAPSHOT.jar # 自己寫的單詞統計,文件放在了hdfs上
這裏注意spark內存使用的配置
遇到的問題
Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources
暫時未解決
集羣配置參考
- http://wuchong.me/blog/2015/04/04/spark-on-yarn-cluster-deploy/
- https://www.jianshu.com/p/aa6f3a366727