Connecting Hortonworks HDP to OBS

Deployment View

Version Information

Hardware: 1 Master + 3 Cores (flavor: 8U32G; OS: CentOS 7.5)

Software: Ambari 2.7.1.0 and HDP 3.0.1.0

Deployment View

Updating OBSA-HDFS

1. Download the OBSA-HDFS that matches the Hadoop version.
   Download the OBSA-HDFS JAR package (for example, hadoop-huaweicloud-3.1.1-hw-53.8.jar) to the /mnt/obsjar directory.

   

   • In a hadoop-huaweicloud-x.x.x-hw-y.jar package name, x.x.x indicates the Hadoop version number, and y indicates the OBSA version number. For example, in hadoop-huaweicloud-3.1.1-hw-53.8.jar, 3.1.1 is the Hadoop version number, and 53.8 is the OBSA version number.
   • If the Hadoop version is 3.1.x, select hadoop-huaweicloud-3.1.1-hw-53.8.jar.

2. Copy the downloaded OBSA-HDFS JAR package to the following directories:

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/share/hst/activity-explorer/lib/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/3.0.1.0-187/hadoop-mapreduce/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/3.0.1.0-187/spark2/jars/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/3.0.1.0-187/tez/lib/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /var/lib/ambari-server/resources/views/work/CAPACITY-SCHEDULER{1.0.0}/WEB-INF/lib/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /var/lib/ambari-server/resources/views/work/FILES{1.0.0}/WEB-INF/lib/

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /var/lib/ambari-server/resources/views/work/WORKFLOW_MANAGER{1.0.0}/WEB-INF/lib/

   ln -s /usr/hdp/3.0.1.0-187/hadoop-mapreduce/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/3.0.1.0-187/hadoop-mapreduce/hadoop-huaweicloud.jar

Adding Configuration Items to the HDFS Cluster

1. Add configuration items in file Custom core-site.xml to the ADVANCED in the HDFS cluster's CONFIGS. These items include fs.obs.access.key, fs.obs.secret.key, fs.obs.endpoint, and fs.obs.impl.
   

   1. fs.obs.access.key, fs.obs.secret.key, and fs.obs.endpoint indicate the AK, SK, and endpoint respectively. Enter the actually used AK/SK pair and endpoint. To obtain them, see Access Keys (AK/SK) and Endpoints and Domain Names, respectively.
   2. Set fs.obs.impl to org.apache.hadoop.fs.obs.OBSFileSystem.

2. Restart the HDFS cluster.

Adding Configuration Items to the MapReduce2 Cluster

1. In the mapred-site.xml file under ADVANCED in CONFIGS of the MapReduce2 cluster, change the value of mapreduce.application.classpath to /usr/hdp/3.0.1.0-187/hadoop-mapreduce/*.
2. Restart the MapReduce2 cluster.

Adding a JAR Package for Connecting Hive to OBS

1. Create the auxlib folder on the Hive Server node:

   mkdir /usr/hdp/3.0.1.0-187/hive/auxlib

2. Save the OBSA-HDFS JAR package to the auxlib folder:

   cp /mnt/obsjar/hadoop-huaweicloud-3.1.1-hw-53.8.jar /usr/hdp/3.0.1.0-187/hive/auxlib

3. Restart the Hive cluster.