Table of Contents:

Prerequisites checklist

Use the checklist to track the prerequisites setup progress at your project.

Hadoop prerequisites

Open ports

In a controlled network environment, it is common to have firewall rules in place. In order to enable communication of SAP systems with Hadoop, the following port numbers should be reachable in the Hadoop cluster from the SAP system:

These are the default port numbers of Hadoop services.

If Kerberos is enabled, KDC (Key Distribution Center) should also be reachable on port 88 (tcp/udp) from each SAP application server.

DNS names

Proper DNS name translation should be configured between SAP and Hadoop for Kerberos communication.
DNS resolution should be tested from SAP host using OS command 'nslookup <hadoop_host_FQDN>'.
If IBM Java is used, also reverse lookup must be successful - IP address to hostname.

Hive parameters

Two configuration parameters of the Hive service must be configured in Hive Service Advanced Configuration Snippet (Safety Valve) for Hive-site.xml.

hive.exec.dynamic.partition = true
hive.exec.dynamic.partition.mode = nonstrict

Example:

NOTE: If Simba JDBC drivers will be used, these parameters can be set for our session only in the SAP system, therefore without a global impact on the cluster.

Hadoop technical user

We recommend creating distinct technical users for every SAP system connected to the Hadoop cluster in order to isolate the system's data.
There is usually a central repository for Hadoop users (LDAP/AD), but users can also be created locally (on every Hadoop cluster node).
Recommended naming convention is reflecting SAP <sid>adm users → <sid>hdp

Each Hadoop technical user should have its own dedicated group, in case Sentry service is used for authorization management as in Sentry, access roles are assigned to groups.

For illustration purposes, we will use Hadoop user dvqhdp (with group dvqhdp) in further text.

If Kerberos is used

Create Kerberos principal in form of <sid>hdp@<KERBEROS_REALM>. This can be either a principal created in MIT Kerberos, FreeIPA, or an Active Directory user. 
To export a Kerberos keytab from the Active directory, use the following command:

ktpass /princ dvqhdp@HADOOP.LOCAL /pass badpassword1 /ptype KRB5_NT_PRINCIPAL /out DVQ.keytab

HDFS landing zone

A landing zone, typically a home folder of the technical user, needs to be created on HDFS. 

The directory needs to meet the following conditions:

• Technical user needs to be able to read and write to this directory and all subdirectories
• Impala and Hive user need to be able to read and write to this directory and all subdirectories

NOTE: If Kerberos is used, Impala and Hive runtime keytabs are stored by Cloudera Manager under /var/run/cloudera-scm-agent/process/ or /etc/security/keytabs in Hortonworks distribution.

Example landing zone creation:

## Create a home directory
[root@skbtscck21 ~]# hadoop fs -mkdir -p /user/dvqhdp/.Trash
## Set ownership and permissions
[root@skbtscck21 ~]# hadoop fs -chown -R dvqhdp:dvqhdp /user/dvqhdp
[root@skbtscck21 ~]# hadoop fs -chmod -R 770 /user/dvqhdp
## Set ACL to grant access to Hive group (by default containing Hive and Impala user)
[root@skbtscck21 ~]# hadoop fs -setfacl -m default:group:hive:rwx /user/dvqhdp
## Check the directory
[root@skbtscck21 ~]# hadoop fs -ls -d /user/dvqhdp
drwxrwx---+  - dvqhdp dvqhdp	0 2017-03-22 14:45 /user/dvqhdp

Hive database

We recommend creating a dedicated database (schema) in Hive for each SAP system. The recommended database name is sap<sid> (e.g.: sapdvq).

Access privileges on the Hadoop side

In the productive Hadoop cluster, depending on Hadoop distribution, either Sentry service or Ranger service is responsible for the management of users' privileges on Hadoop resources.
For proper functionality of Storage Management, Hadoop technical user <sid>hdp needs to have access to the following two resources at least:

1. User's HDFS home directory - typically /user/<sid>hdp
2. Assigned Hive database - typically sap<sid>

To set up necessary policies in the respective security service, follow the instructions below.

If Sentry is used

Sentry service is managing access to Hadoop resources using Sentry rules. The rules are created for a role, which can have "one to many" relations with user groups (not the users directly).
We typically set up only "one to one" role ↔ group relations. You need to set up two rules, one granting ALL actions on the HDFS directory (gets automatically translated to URI) and one granting all actions on the Hive database.

Example:

NOTE: If HDFS ACL synchronization with Sentry rules is enabled, add the user's directory or any parent directory to the "Sentry Synchronization Path Prefixes" parameter in the HDFS service configuration.
More information on the HDFS ACL synchronization topic can be found in the chapter Synchronizing HDFS ACLs and Sentry Permissions.

If Ranger is used

Similar to Sentry, the Ranger service is managing access to Hadoop resources using policies. The policies can grant privileges either to the group or to users directly.
Again, you need to create two policies - one granting full access to Hadoop technical user on his HDFS directory and the other one granting access to the Hive database.

HDFS policy example:

Hive database policy example:

Verification of setup

To verify the setup on the Hadoop side is valid, we recommend using the technical user to create a testing table inside the sap<sid> Hive database.
Load a test file to the user's home directory, verify whether it's there and delete it in the end.
If all the commands are successful, you can conclude that the setup is valid.

[root@skbtscck21 ~]# database=sapnsq	\\set the database variable
[root@skbtscck21 ~]# jdbc_string="jdbc:hive2://skbtscck21.hadoop.local:10000/$database;principal=hive/skbtscck21.hadoop.local@HADOOP.LOCAL;ssl=true;sslTrustStore=/opt/certs/jks/skbtscck21-keystore.jks;trustStorePassword=123456aB"	\\ set jdbc string
[root@skbtscck21 ~]# kinit -kt ~/nsqhdp.keytab nsqhdp@HADOOP.LOCAL	\\ authenticate as the technical user
[root@skbtscck21 ~]# klist	\\ verify that you have a valid ticket
Ticket cache: FILE:/tmp/krb5cc_0
Default principal: nsqhdp@HADOOP.LOCAL

Valid starting       Expires              Service principal
04/25/2019 17:24:03  04/26/2019 17:24:03  krbtgt/HADOOP.LOCAL@HADOOP.LOCAL
        renew until 05/02/2019 17:24:03
[root@skbtscck21 ~]# beeline -u $jdbc_string -e "create table if not exists $database.xxx (a int)"	\\ Create test table
INFO  : OK                                                                                                         
[root@skbtscck21 ~]# beeline -u $jdbc_string -e "drop table $database.xxx"	\\ delete test table
INFO  : OK
[root@skbtscck21 ~]# beeline -u $jdbc_string -e "show current roles"	\\ only with Sentry - show roles of the user
INFO  : OK
+---------------+--+
|     role      |
+---------------+--+
| sapnsq_admin  |
+---------------+--+
[root@skbtscck21 ~]# beeline -u $jdbc_string -e "show grant role sapnsq_admin"	\\ only with Sentry - display privileges of the role. Should look like the one below
INFO  : OK
+--------------------------------------------------+--------+------------+---------+-----------------+-----------------+------------+---------------+-------------------+----------+--+
|                     database                     | table  | partition  | column  | principal_name  | principal_type  | privilege  | grant_option  |    grant_time     | grantor  |
+--------------------------------------------------+--------+------------+---------+-----------------+-----------------+------------+---------------+-------------------+----------+--+
| sapnsq                                           |        |            |         | sapnsq_admin    | ROLE            | all        | false         | 1490282925235000  | --       |
| hdfs://skbtscck21.hadoop.local:8020/user/nsqhdp  |        |            |         | sapnsq_admin    | ROLE            | all        | false         | 1490282925256000  | --       |
+--------------------------------------------------+--------+------------+---------+-----------------+-----------------+------------+---------------+-------------------+----------+--+

[root@skbtscck21 ~]# echo Success > file.txt	\\ Create a test file
[root@skbtscck21 ~]# hadoop fs -copyFromLocal ./file.txt /user/nsqhdp/	\\ Store file on HDFS
[root@skbtscck21 ~]# hadoop fs -cat /user/nsqhdp/file.txt	\\ Read file
Success
[root@skbtscck21 ~]# hadoop fs -rm -skipTrash /user/nsqhdp/file.txt	\\ Clean up
Deleted /user/nsqhdp/file.txt

OS prerequisites (On SAP host)

This group of requirements relates to the operating systems underlying the SAP system with all its application servers. SNP products (e.g. SNP Glue™, SNP OutBoard™ Data Tiering) have been developed and tested on the SUSE Linux environment and Windows Server 2012.
By design, they are not limited by the choice of an operating system, if the requirements listed in this guide are met. Successful implementations were done also on AIX and Solaris.

OS directories

Hadoop connector uses two directories used to store configuration and log files.
Create them with appropriate permissions (read/write) and usual SAP directory ownership (<sid>adm:sapsys).

1. dvd_conn directory in shared /sapmnt filesystem:

   $ ls -ld /sapmnt/DVQ/global/security/dvd_conn
   drwx------ 2 dvqadm sapsys 4096 --- /sapmnt/DVQ/global/security/dvd_conn
   

   This one contains drivers, Kerberos, and SSL-related files. It is shared among all SAP application servers.

2. dvd_conn directory in the work directory of each SAP application server:
   

   $ ls -ld /usr/sap/DVQ/DVEBMGS05/work/dvd_conn
   drwxr-xr-x 7 dvqadm sapsys 4096 --- /usr/sap/DVQ/DVEBMGS05/work/dvd_conn

   This one stores Java Connector libraries, configuration, and log files. The folder can be also created automatically during Java connector setup in /DVD/JCO_MNG. 

$ ls -ld /usr/sap/DVQ/dvd_conn
drwxr-xr-x 2 dvqadm sapsys 4096 --- /usr/sap/DVQ/dvd_conn

JDBC Drivers

JDBC protocol is used to connect to Hadoop services (Hive and Impala). JDBC drivers have to be manually stored on the operating system and be accessible to the connector.
We recommend storing the drivers in the shared dvd_conn directory, organized in sub-directories to avoid possible conflicts.

$ ls -ld /sapmnt/DVQ/global/security/dvd_conn/[hi]*
drwxr-x--- 2 dvqadm sapsys 4096 --- /sapmnt/DVQ/global/security/dvd_conn/hive
drwxr-x--- 3 dvqadm sapsys 4096 --- /sapmnt/DVQ/global/security/dvd_conn/impala

JDBC drivers can be usually downloaded from your Hadoop vendor's website.
Java connector is currently compiled with Java 7, which is directly compatible with Cloudera JDBC drivers version 41.
For the future, JDBC42 is recommended to be used with Java 8 (but we are not there yet).

There are multiple drivers able to facilitate communication between Java Connector and respective Hadoop service.
Based on our experience, the most reliable are Simba drivers (adopted by Cloudera, Hortonworks, and MapR Hadoop distributions).

Kerberos keytab and configuration files

The Kerberos keytab of <sid>hdp principal should be exported from the Kerberos database, copied into the operating system directory /sapmnt/<SID>/global/security/dvd_conn, and made available to the <sid>adm user:

/sapmnt/DVQ/global/security/dvd_conn # ls -l DVQ.keytab
-r-------- 1 dvqadm sapsys 59 Apr  5 15:12 DVQ.keytab

At the same location, the Kerberos configuration file should be created or copied and made readable for the user <sid>adm.

Usually a suitable krb5.conf file can be found on Hadoop nodes. Here is a sample of the Kerberos configuration file:

/sapmnt/DVQ/global/security/dvd_conn # ls -l krb5.conf
-r-------- 1 dvqadm sapsys 393 Feb 22 16:28 krb5.conf

/sapmnt/DVQ/global/security/dvd_conn # cat krb5.conf
[libdefaults]
default_realm = HADOOP.LOCAL
dns_lookup_kdc = false
dns_lookup_realm = false
ticket_lifetime = 86400
renew_lifetime = 604800
forwardable = true
default_tgs_enctypes = rc4-hmac
default_tkt_enctypes = rc4-hmac
permitted_enctypes = rc4-hmac
udp_preference_limit = 1
kdc_timeout = 3000
[realms]
HADOOP.LOCAL = {
kdc = hadoop01.hadoop.local
admin_server = hadoop01.hadoop.local
}

Kerberos keytab verification

To verify that Kerberos config and keytab are valid, execute the following steps.

## 	switch to <sid>adm
vsks012:~ # su - dvqadm
## change to dvd_conn directory
vsks012:dvqadm 54> cd /sapmnt/DVQ/global/security/dvd_conn/	
## check principal in the keytab
vsks012:dvqadm 55> klist -k DVQ.keytab		
Keytab name: FILE:DVQ.keytab
KVNO Principal
---- --------------------------------------------------------------------------
   9 dvqhdp@DATA.DEV

## Set environment variable for kerberos config and login as the technical user. This is the expected result
vsks012:dvqadm 56> setenv KRB5_CONFIG /sapmnt/DVQ/global/security/dvd_conn/krb5.conf && kinit -kt DVQ.keytab dvqhdp@DATA.DEV && klist		 
Ticket cache: FILE:/tmp/krb5cc_1001
Default principal: dvqhdp@DATA.DEV

Valid starting     Expires            Service principal
05/14/19 11:04:08  05/15/19 11:04:08  krbtgt/DATA.DEV@DATA.DEV
        renew until 05/21/19 11:04:08

Kerberos 4 ticket cache: /tmp/tkt1001
klist: You have no tickets cached

SSL Certificates for Java

If your Hadoop cluster has full SSL/TLS encryption enabled, it’s necessary to create a Java trust store and save it on the following path with correct ownership and permissions:

/sapmnt/<SID>/global/security/dvd_conn # ls -l jssecacerts
-r-------- 1 <SID>adm sapsys 59 Apr  5 15:12 jssecacerts

This trust store needs to contain the CA certificate with which Hadoop host certificates were signed. Another possibility is to store the respective hosts' self-signed certificates.

An alternative option is to copy the complete jssecacerts trust store from any Hadoop node and place it in this path.

SAP prerequisites

Kerberos cookie encoding

By default, the SAP system encodes certain characters in cookies. SAP Note 1160362 describes the behavior in more detail. As the Kerberos cookie must not be anyhow modified for the Kerberos server to accept it, this encoding should be disabled by setting the following parameter in each SAP application server's instance profile:

ict/disable_cookie_urlencoding = 1

ict/disable_cookie_urlencoding = 2

The parameter is dynamic in kernel version 7.53 and higher.

SSL for SAP RFCs

To enable SSL communication for SAP RFCs, add client certificates enabling communication with Hadoop nodes to the SAP certificate list in the transaction STRUST.

If HttpFS is used, a client certificate of the HttpFS host is required.

If WebHDFS is used, client certificates of all datanodes and namenodes are necessary.

To import necessary certificates:

1. Use transaction STRUST.
2. In the left menu choose the certificate list that you want to add the certificate to (by double-clicking).
3. In the right window area in the bottom left, click Import.
4. In tab FILE of the dialog window point to the certificate file on your local file system (certificate should be in .pem format).
5. After you confirm the path and SAP is able to recognize the certificate, details are displayed in the corresponding fields.
6. To complete adding the certificate click Add to Certificate List.
7. Click Save (in the general menu or Ctrl+S).
8. Restart ICM in transaction SMICM: Menu > Administration > ICM > Exit Soft > Local/Global.

Check the following parameters (typically they are set and active):

HTTP service and ICM parameters

The HTTP service must be active in the SAP system. It can be checked via transaction SMICM > [Goto] > Services

If the HttpFS/WebHDFS service is SSL secured, the HTTPS service needs to be active as well.

The following two parameters are affecting HTTP communication of the SAP system. By default they do not need to be modified and are listed here just for the information purpose:

• icm/HTTP/client/keep_alive_timeout – HTTP communication timeout, can be raised in the case of HTTP communication failing in timeout.
• icm/HTTP/max_request_size_KB – maximum size of data which ICM accepts (default 100 MB).

Java connector

Java connector is a critical middle-ware component. Please follow the steps in this guide Java Connector Setup to set it up before you continue. 

Configuration

When all prerequisites are fulfilled, further configuration is performed in the SAP system.

RFC Destination 

An RFC destination that needs to be created via transaction SM59.

HttpFS/WebHDFS RFC

Storage management communicates with HDFS using WebHDFS API. In the reference configuration, we will be using the HttpFS service, but this can be substituted with WebHDFS if necessary.
The only difference in the setup when WebHDFS is used is the port number and SSL certificates required in STRUST.

This RFC connection is used for communication with Hadoop's HttpFS service which mediates operations in HDFS.
Two RFCs pointing to different HttpFS services can be created to ensure High Availability.

The name and description of the destination are optional, but it is recommended to designate its purpose with the keywords 'Hadoop' and 'HttpFS'. In our example, the RFC destination also contains the Hadoop server hosting the HttpFS service for the sake of clarity:

Entries explained:

• Connection Type – G for HTTP connection to an external service
• Target host – FQDN of Hadoop server hosting HttpFS service
• Service No. – Port number on which HttpFS service is listening (14000 for HttpFS, 50070 for WebHDFS)
• Path Prefix – This string consists of two parts:
  1. /webhdfs/v1 part is mandatory
  2. /user/dvqhdp part defines the Hadoop user's 'root' directory in HDFS where flat files from the SAP system are loaded

If SSL is used: Enable SSL and add client certificate PSE which is used in Logon & Security tab.

Logical paths and logical filenames

SNP ships default logical paths leading to files stored on SAP application servers. These default values are usable for most SAP installations and in general, don't need to be changed. If these default paths are used, the administrator needs to make sure that all security-related files (Kerberos keytab, Kerberos config, jssecacerts) and drivers stick to the default naming used in these paths. 

In shipped SNP paths, standard SAP logical paths DIR_HOME and DIR_GLOBAL are frequently used as variables. These paths usually refer to DIR_HOME = /usr/sap/<SID>/<INSTANCE>/work and DIR_GLOBAL = /sapmnt/<SID>/global/

Logical files and their values shipped by SNP are the following:

• /DVD/DEF_KRB_KEYTAB = <P=DIR_GLOBAL>/security/dvd_conn/<SID>.keytab
• /DVD/DEF_KRB_CONFIG = <P=DIR_GLOBAL>/security/dvd_conn/krb5.config
• /DVD/DEF_SSL_TRUSTSTORE = <P=DIR_GLOBAL>/security/dvd_conn/jssecacerts
• /DVD/DEF_JCO_DIR = <P=DIR_HOME>/dvd_conn/
• /DVD/DEF_HIVE_DRIVER = <P=DIR_GLOBAL>/security/dvd_conn/hive
• /DVD/DEF_IMPALA_DRIVER = <P=DIR_GLOBAL>/security/dvd_conn/impala

If a customer wants to use a custom location or filename, a Z copy of these logical paths and logical files with customizations needs to be made, as any direct change to these paths would be overwritten during the Storage Management update.

Storage Management setup

Storage Management facilitates transparent communication with different types of storages, HDFS for flat files, and Hive for structured data.
In order to transparently store data, you two types of Hadoop storages need to be defined in Storage Management (/DVD/SM_SETUP):

• HDFS storage which facilitates the transfer of files to HDFS through the Hadoop HttpFS service
• Hive storage which enables data replication between SAP tables and Hive tables

The third type of storage is required for efficient querying of data located in Hive:

• Impala storage connects to Impala agents to provide fast SQL execution by leveraging Impala in-memory data caching

HDFS storage type

Sample entry:

Entries explained:

• HTTP RFC Destination – HttpFS RFC destination defined in SM59
• HTTP RFC Destination (HA) – Secondary HttpRFC destination defined in SM59 as a failover
• HDFS Repeat – Number of retries if request to HDFS is not successful
• HDFS Repeat Time (seconds) – Delay between request attempts
  
• HDFS Permissions – Files will be created with these permissions (e.g. 770 for -rwxrwx---)

Authentication settings:

• Authentication method – Authentication method for the Hadoop cluster
• Username – Hadoop user-principal 
• JCO RFC destination – Authentication RFC destination created in Authentication RFC
• Krb. Config file path – Logical file definition for Kerberos configuration file, default /DVD/DEF_KRB_CONFIG can be used
• Krb. Keytab path – Logical file definition for Kerberos keytab file, default /DVD/DEF_KRB_KEYTAB can be used
• SSL Keystore – Logical file definition for SSL Keystore, default /DVD/DEF_SSL_TRUSTSTORE can be used
• SSL Password – Password for accessing SSL Keystore (in edit mode should be written in lower line and [Hash]-ed)
  

Complete the creation of the storage by confirming (F8). If the SAP system can authenticate against Hadoop Kerberos and get properties of HDFS home directory (/user/<sid>hdp) from the HttpFS service, storage creation is considered successful.

HIVE storage type

The Hive metastore storage is created in a similar way as the process of setting up the HDFS storage, but the values are different:

Finish the creation of the storage by confirming (F8). If the SAP system is able to authenticate against Hadoop Kerberos and receives the expected result of the SQL command 'use database', the creation of the storage is considered successful.

Extended explanation for Storage setup with Hadoop distribution other than Cloudera

The following fields' values can differ depending on the Hadoop distribution used:

• Load driver classname – Apart from Cloudera, there are other vendors offering JDBC drivers able to connect to Hive service (or other Hadoop services like Impala or Drill).
  In the case of Hortonworks distribution, the JDBC Hive driver can be found directly on the Hadoop node among other JAR files at /usr/hdp/current/hadoop-client/client/.

  The contents of the jar file can be listed with the 'less' command:

  [root@skbtshcc01 ~]# less /usr/hdp/2.6.4.0-91/hive2/jdbc/hive-jdbc-2.1.0.2.6.4.0-91-standalone.jar | grep HiveDriver
  -rw---- 2.0 fat 6960 bl defN 18-Jan-04 10:39 org/apache/hive/jdbc/HiveDriver.class

  The driver's Class Name can be derived from it as

  org.apache.hive.jdbc.HiveDriver

  
  

  Based on experience, hive-jdbc-xxx-standalone.jar is not as "standalone" as the name would suggest.
  Three other JAR files are needed in the same driver directory when the connection is established via Zookeeper (these may vary depending on the Hadoop version):hadoop-auth-X.X.X.X.X.X.X-XX.jar
  hadoop-common-X.X.X.X.X.X.X-XX.jar
  hadoop-yarn-registry-X.X.X.X.X.X.X-XX.jar

  Reference links:
  https://www.cloudera.com/downloads/connectors/hive/jdbc/2-6-5.html
  https://www.simba.com/resources/jdbc/
  https://www.progress.com/jdbc/apache-hadoop-hive
  http://repo.hortonworks.com/content/repositories/releases/org/apache/hive/hive-jdbc/

• Custom connection string – Not used under standard Cloudera setup, as the URL connection string is being composed of other settings.
  In case of a specific configuration on the Hadoop side or for testing purposes, it is possible to use an explicitly stated connection URL here.
  Example: connection URL directed at Zookeeper service (port 2181) running on three Hadoop hosts, acting as proxy redirecting connection to currently active Hive server in High Availability setup

  jdbc:hive2://skbtshcc01.hadoop.local:2181,skbtshcc03.hadoop.local:2181,skbtshcc02.hadoop.local:2181/;serviceDiscoveryMode=zooKeeper;zooKeeperNamespace=hiveserver2
  
  

  Reference links: 
  https://cwiki.apache.org/confluence/display/Hive/HiveServer2+Clients
  https://www.cloudera.com/documentation/other/connectors/hive-jdbc/latest/Cloudera-JDBC-Driver-for-Apache-Hive-Install-Guide.pdf
  https://docs.microsoft.com/bs-latn-ba/azure/hdinsight/hadoop/apache-hadoop-connect-hive-jdbc-driver?view=aspnetcore-2.1
  https://docs.datafabric.hpe.com/51/Hive/ConnectingtoHiveServer2-U_29655382-d3e110.html
  
  
• File type – Format in which the files loaded to HDFS is stored. It depends on Hadoop distribution - in Cloudera and MapR it's typically PARQUET, in Hortonworks it's typically ORC.
  
  
• Compression codec – Optional setting. It depends on the previously selected File Type and whether the files stored in HDFS should be compressed or not. If the file type is PARQUET, the compression codec is typically SNAPPY. If the file type is ORC, the compression codec is typically ZLIB.
  
  
• Use Cloudera drivers – Depending on this checkbox, the default connection URL is generated either with JDBC Cloudera/Simba driver syntax (checkbox ticked) or with JDBC Apache Hive driver syntax (checkbox unticked).
  If a custom connection string is used, this should be left unchecked.