一、 安裝PXF3.3.0.0，這裡所安裝的pxf的包檔案都包含在apache-hawq-rpm-2.3.0.0-incubating.tar.gz裡面
下面步驟都是以root身份執行
這裡注意，pxf外掛要用到tomcat服務，必須使用安裝包裡面的7.0.62， 不能安裝或升級為 tomcat8，這會造成依賴的catalina.jar包 版本不匹配，以致pxf啟動！

安裝時由於pxf的包都是el6版本的，但是我用的centos7，所以rpm帶引數“--nodeps”以忽略RPM命令對依賴包的檢測。

cd /opt/gpadmin/hawq_rpm_packages

rpm -ivh  apache-tomcat-7.0
 
.62-el6.noarch.rpm 
rpm -ivh --nodeps pxf-service-3.3.0.0-1.el6.noarch.rpm
rpm -ivh --nodeps pxf-hdfs-3.3.0.0-1.el6.noarch.rpm 
rpm -ivh --nodeps pxf-hive-3.3.0.0-1.el6.noarch.rpm 
rpm -ivh --nodeps pxf-hbase-3.3.0.0-1.el6.noarch.rpm 
rpm -ivh --nodeps pxf-jdbc-3.3.0.0-1.el6.noarch.rpm 
rpm -ivh --nodeps pxf-json-3.3
 
.0.0-1.el6.noarch.rpm
rpm -ivh --nodeps pxf-3.3.0.0-1.el6.noarch.rpm

二、 配置PXF

1，由於hadoop為HDP版本，所以使用hdp相關的jar包配置

[[email protected] ~]cp  /etc/pxf/conf/pxf-privatehdp.classpath  /etc/pxf/conf/pxf-private.classpath
[[email protected] ~]cp  /etc/pxf/conf/pxf-profiles.xml   /etc/pxf/conf/pxf-profiles.default.xml

2，修改pxf目錄所有者：

[[email protected] ~] chown -R pxf:pxf /opt/pxf-3.3.0.0
[[email protected]-bd01 ~] chown -R pxf:pxf /tmp/logs

3，建立軟連線目錄：

[[email protected]-bd01 ~] ln -s /etc/pxf-3.3.0.0/conf /opt/pxf-3.3.0.0/conf
[[email protected]-bd01 ~] ln -s /usr/lib/pxf  /opt/pxf-3.3.0.0/lib

4，建立一個init過程需要的目錄和template檔案

[[email protected] ~] mkdir /opt/pxf-3.3.0.0/conf-templates
[[email protected]-bd01 ~] cp /opt/pxf/conf/pxf-privatehdp.classpath /opt/pxf/conf-templates/pxf-private-hdp.classpath.template

5，修改pxf-service檔案為pxf，因為init過程需要建立同名的目錄，同時需要修改init.d目錄中的連結：

[[email protected] ~] mv /opt/pxf/pxf-service  /opt/pxf/pxf
[[email protected]-bd01 ~] unlink /etc/init.d/pxf-service 
[[email protected]-bd01 ~] ln -s /opt/pxf/pxf   /etc/init.d/pxf-service

6，修改tomcat/conf目錄的許可權，同時在pxf目錄中建立連結

[[email protected] ~] chmod 755 -R /opt/apache-tomcat/conf/
[[email protected]-bd01 ~] ln -s /opt/apache-tomcat /opt/pxf-3.3.0.0/apache-tomcat

7，編輯/etc/pxf/conf/pxf-env.sh，修改 PARENT_SCRIPT_DIR和LD_LIBRARY_PATH的值

[[email protected] ~] vim /etc/pxf/conf/pxf-env.sh 
export PARENT_SCRIPT_DIR=/opt/pxf-3.3.0.0
export PXF_HOME=/opt/pxf-3.3.0.0
export LD_LIBRARY_PATH=/usr/hdp/current/hadoop-client/lib/native:${LD_LIBRARY_PATH}

8，修改pxf指令碼檔案，設定PXF_HOME

[[email protected] ~] vim /opt/pxf-3.3.0.0/pxf

export PARENT_SCRIPT_DIR=/opt/pxf-3.3.0.0

export PXF_HOME=/opt/pxf-3.3.0.0

9，編輯/etc/pxf/conf/pxf-public.classpath，新增一系列的jar包

[[email protected] pxf-3.3.0.0]# vim /etc/pxf-3.3.0.0/conf/pxf-public.classpath
/usr/hdp/current/hadoop-client/lib/commons-beanutils-1.9.3.jar
/usr/hdp/current/hadoop-client/lib/commons-cli-1.2.jar
/usr/hdp/current/hadoop-client/lib/commons-codec-1.11.jar
/usr/hdp/current/hadoop-client/lib/commons-collections-3.2.2.jar
/usr/hdp/current/hadoop-client/lib/commons-compress-1.4.1.jar
/usr/hdp/current/hadoop-client/lib/commons-configuration2-2.1.1.jar
/usr/hdp/current/hadoop-client/lib/commons-io-2.5.jar
/usr/hdp/current/hadoop-client/lib/commons-lang-2.6.jar
/usr/hdp/current/hadoop-client/lib/commons-lang3-3.4.jar
/usr/hdp/current/hadoop-client/lib/commons-logging-1.1.3.jar
/usr/hdp/current/hadoop-client/lib/commons-math3-3.1.1.jar
/usr/hdp/current/hadoop-client/lib/commons-net-3.6.jar
/usr/hdp/current/hadoop-client/lib/jersey-core-1.19.jar
/usr/hdp/current/hadoop-client/lib/jersey-json-1.19.jar
/usr/hdp/current/hadoop-client/lib/jersey-server-1.19.jar
/usr/hdp/current/hadoop-client/lib/jersey-servlet-1.19.jar
/usr/hdp/current/hadoop-client/lib/jsr311-api-1.1.1.jar
/usr/hdp/current/hadoop-client/lib/woodstox-core-5.0.3.jar
/usr/hdp/current/hadoop-client/lib/stax2-api-3.1.4.jar
/usr/hdp/current/hadoop-client/lib/htrace-core4-4.1.0-incubating.jar
/usr/hdp/current/hadoop-client/lib/re2j-1.1.jar
/usr/hdp/3.0.0.0-1634/hbase/lib/atlas-hbase-plugin-impl/commons-configuration-1.10.jar
/usr/hdp/current/hadoop-hdfs-datanode/hadoop-hdfs-rbf.jar
/usr/hdp/current/hadoop-hdfs-datanode/hadoop-hdfs-nfs.jar
/usr/hdp/current/hadoop-hdfs-datanode/hadoop-hdfs-native-client.jar
/usr/hdp/current/hadoop-hdfs-datanode/hadoop-hdfs-httpfs.jar
/usr/hdp/current/hadoop-hdfs-datanode/hadoop-hdfs-client-3.1.0.3.0.0.0-1634.jar
/opt/pxf-3.3.0.0/lib/pxf-service-3.3.0.0.jar
/opt/pxf-3.3.0.0/lib/pxf-api-3.3.0.0.jar
:wq

 10，複製pxf-profiles.xml 為pxf-profiles-default.xml並編輯，新增profile配置

[[email protected] pxf-3.3.0.0]# cp /etc/pxf/conf/pxf-profiles.xml   /etc/pxf/conf/pxf-profiles-default.xml 
[[email protected]-bd05 pxf-3.3.0.0]# vim /etc/pxf/conf/pxf-profiles-default.xml 
<profiles>
  <profile>
        <name>HdfsTextSimple</name>
        <description>This profile is suitable for using when reading delimited single line records from plain text files
            on HDFS
        </description>
        <plugins>
            <fragmenter>org.apache.hawq.pxf.plugins.hdfs.HdfsDataFragmenter</fragmenter>
            <accessor>org.apache.hawq.pxf.plugins.hdfs.LineBreakAccessor</accessor>
            <resolver>org.apache.hawq.pxf.plugins.hdfs.StringPassResolver</resolver>
        </plugins>
    </profile>
    <profile>
        <name>HdfsTextMulti</name>
        <description>This profile is suitable for using when reading delimited single or multi line records (with quoted
            linefeeds) from plain text files on HDFS. It is not splittable (non parallel) and slower than HdfsTextSimple.
        </description>
        <plugins>
            <fragmenter>org.apache.hawq.pxf.plugins.hdfs.HdfsDataFragmenter</fragmenter>
            <accessor>org.apache.hawq.pxf.plugins.hdfs.QuotedLineBreakAccessor</accessor>
            <resolver>org.apache.hawq.pxf.plugins.hdfs.StringPassResolver</resolver>
        </plugins>
    </profile>
</profiles>

三、初始化pxf，必須使用pxf使用者

1,設定pxf的密碼

passwd pxf

2，初始化，需要使用使用者pxf

[[email protected] pxf]# source /etc/pxf/conf/pxf-env.sh
[[email protected] pxf]# sudo -u pxf service pxf-service init

Generating /opt/pxf-3.3.0.0/conf/pxf-private.classpath file from /opt/pxf-3.3.0.0/conf-templates/pxf-private-hdp.classpath.template ...

cp /opt/pxf/pxf-service/webapps/pxf/WEB-INF/lib/*.jar  /opt/pxf/lib/

四、啟動PXF service

1，啟動：

sudo -u pxf service pxf-service start

Checking if tomcat is up and running...
tomcat not responding, re-trying after 1 second (attempt number 1)
Checking if PXF webapp is up and running...
PXF webapp is listening on port 51200

 2，測試：

使用pxf外掛訪問已經事先從oracle匯入到HDFS上的資料（使用了sqoop的--compress選項，是gz壓縮格式，但是HdfsTextSimple可以直接訪問），下面是建立hawq外部表的命令，注意路徑中的星號。

drop external table ext.yx_bw;
create external table ext.yx_bw (occur_time date, ......    ) 
location ('pxf://192.168.58.15:51200/var/data/ext/yx_bw/*?profile=hdfstextsimple') format 'text'(delimiter ',' null '');    

**注意** ，此處的主機地址，我直接使用的是主機的地址，如果使用主機名稱則hawq訪問失敗，據我觀察應該是沒有正確轉換，一直沒能解決此問題，如果哪位大俠知道請一定不吝賜教，先謝過了！如果地址使用location ('pxf://bd05:51200/var/data/ext/yx_bw/*?profile=hdfstextsimple') ，外部表可以建立，但是訪問資料時顯示如下錯誤，且沒有詳細資訊，pxf服務的log也找不到訪問失敗的記錄！

epbd=> select * from ext.yx_bw;      
ERROR:  remote component error (0): (libchurl.c:897)

下面是系統中的/etc/hosts檔案和/etc/host.conf檔案，由於本叢集可以訪問外網，可以看到nslookup返回了錯誤地址，但是ping和curl訪問都是正確的。

[email protected] pg_log]# cat /etc/hosts
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
192.168.58.11  ep-bd01 bd01
192.168.58.12  ep-bd02 bd02
192.168.58.13  ep-bd03 bd03
192.168.58.14  ep-bd04 bd04
192.168.58.15  ep-bd05 bd05

[[email protected] pg_log]# cat /etc/host.conf 
multi off

[[email protected]bd05 pg_log]# nslookup bd01
Server:         211.137.160.5
Address:        211.137.160.5#53

Non-authoritative answer:
Name:   bd01
Address: 211.137.170.246

[[email protected]-bd05 pg_log]# ping bd01    
PING ep-bd01 (192.168.58.11) 56(84) bytes of data.
64 bytes from ep-bd01 (192.168.58.11): icmp_seq=1 ttl=64 time=0.156 ms
64 bytes from ep-bd01 (192.168.58.11): icmp_seq=2 ttl=64 time=0.160 ms
^C
--- ep-bd01 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 999ms
rtt min/avg/max/mdev = 0.156/0.158/0.160/0.002 ms
[[email protected]-bd05 pg_log]# curl http://bd01:51200/pxf/v0
Wrong version v0, supported version is v15
[roo[email protected] pg_log]# 

五、可選操作：

1，修改/opt/pxf/pxf-service/conf/catalina.properties，修改 base.shutdown.port 

#base.shutdown.port=-1
base.shutdown.port=8005

2，修改/opt/pxf/pxf-service/conf/tomcat-users.xml,給使用者tomcat新增角色manager-gui ，以便可以在瀏覽器中管理webapps

[[email protected] ~] vim /opt/pxf/pxf-service/conf/tomcat-users.xml 
<role rolename="tomcat"/>
<role rolename="manager-gui"/>
<user username="tomcat" password="tomcat" roles="tomcat,manager-gui"/>
<user username="both" password="tomcat" roles="tomcat,role1"/>
<user username="role1" password="tomcat" roles="role1"/>