Architecture 

For Datavard specific scenarios, we need to handle communication with:

1. HDFS 
2. Hive/Impala

HDFS

HDFS (Hadoop Distributed File System) is data storage system used by Hadoop applications, and also Datavard uses it for data transfer. 

It is possible to send files directly to HDFS, but also to use it as temporary storage when transferring data from SAP to Hadoop engines.

Communication with HDFS

Communication with HDFS is provided via WebHDFS REST API which is handled by Datavard storage management directly.

Hive/Impala

Apache Hive is data warehouse software in Hadoop ecosystem, that facilitates reading, writing, and managing large datasets in distributed storage using SQL-like query language.

Apache Impala is massively parallel processing (MPP) SQL query engine for data stored in Hadoop cluster. Impala brings scalable parallel database technology, enabling users to issue low-latency SQL queries to data stored in HDFS.

For purpose of data transfer to Hive/Impala engines, also HDFS is used. Firstly, the data is moved in form of .csv file to HDFS, and afterwards the engine loads the transferred data. 

Communication with Hive/Impala

When communicating with Hive/Impala engines, Java connector (implemented by Datavard) is used. This connector wraps SQL like queries using JDBC jars and forwards them to the engines themselves.

Setup Hadoop storage

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 need to be reachable in Hadoop cluster from the SAP system:

These are default port numbers of Hadoop services.

If Kerberos is enabled, also KDC (Key Distribution Center) needs to be reachable from SAP on port 88 (tcp/udp).

DNS names

Proper DNS name translation needs to be configured between SAP and Hadoop for Kerberos communication.

Hive parameters

There are two configuration parameters of the Hive service, which need to be properly 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:

Hadoop user

We recommend to create distinct users for every SAP system connected to the Hadoop cluster in order to isolate each system's data. 
There is usually a central repository for Hadoop users (LDAP/AD), but you can also create the user locally (on every Hadoop cluster node).

Each of these users should have its own dedicated user group.
If Hadoop Sentry is used: User groups will be used for the definition of Sentry access rules.

The recommended user names are mirroring SAP's guideline for user names: <sid>adm => <sid>hdp.

Create the user's Kerberos principal in form of <sid>hdp@<KERBEROS_REALM>.
The user's home directory on HDFS must be created manually with appropriate permissions:

$ hdfs dfs -ls -d /user/dvqhdp
-rwxrwxr-x 3 dvqhdp supergroup 0 --- /user/dvqhdp

If Kerberos is used: Create Kerberos principal for the user. You need to run kadmin.local on the host, where Kerberos DB is running.

Hive database

Datavard recommends creating a dedicated database (schema) in Hive for each SAP system. The recommended database name is sap<sid> (sapdvq).

If Hadoop Sentry is used: Two Sentry rules have to be created, which enable all actions of <sid>hdp user on sap<sid> database and his home directory in HDFS.
Example: If HDFS ACL synchronization with Sentry permissions is enabled, the user's directory must be added to the Sentry Synchronization Path Prefixes parameter in the HDFS service configuration.
More on HDFS ACL synchronization topic can be found on https://www.cloudera.com/documentation/enterprise/latest/topics/sg_hdfs_sentry_sync.html

OS prerequisites (On SAP host)

This group of requirements relates to the operating systems underlying SAP system with all its application servers. Datavard products (e.g. Datavard Glue, OutBoard DataTiering) has been developed and tested on the SUSE Linux environment, but by design is not limited by the choice of operating system, if requirements listed in this guide are met.

Java version

An up to date Java version (recommended is 1.7 or higher) needs to be available to the SAP's <sid>adm user.

If Hadoop cluster has Kerberos enabled, patched version of Java is required (e.g. Java 8u172).

SAP Java Connector library

SAP Java Connector 3.0 library libsapjco3.so needs to be accessible, which is available for download from the SAP marketplace. It needs to be present in the SAP kernel directory or in other directory pointed by LD_LIBRARY_PATH environment variable of <sid>adm user.

$ which libsapjco3.so
/usr/sap/DVQ/SYS/exe/uc/linuxx86_64/libsapjco3.so

OS directories

Hadoop connector uses two directories dedicated to its configuration and log files:

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

The first one (/sapmnt/<SID>/global/security/dvd_conn) is used to store Kerberos and SSL related files and is shared among SAP application servers.

The second will store drivers, configuration and log files of Java connector used by Datavard Glue. Ownership and permissions need to be set appropriately to <sid>adm

Create directories on each SAP application server to store Storage Management related configuration and log files. <sid>adm user needs permissions to access those folders.

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 accessible to Datavard connector.

Datavard recommends storing the drivers in a folder within the connector directory, organized in subfolders to avoid possible conflicts.

$ ls -ld /usr/sap/DVQ/dvd_conn/drivers/hive* /usr/sap/DVQ/dvd_conn/drivers/impala*
drwxr-xr-x 2 dvqadm sapsys 4096 --- /usr/sap/DVQ/dvd_conn/drivers/hive_apache
drwxr-xr-x 3 dvqadm sapsys 4096 --- /usr/sap/DVQ/dvd_conn/drivers/hive_cloudera
drwxr-xr-x 3 dvqadm sapsys 4096 --- /usr/sap/DVQ/dvd_conn/drivers/impala_cloudera

Kerberos keytab and configuration files

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

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

At the same location, the Kerberos configuration file needs to be created/copied and made readable for the user <sid>adm.
Here is a sample of Kerberos configuration file:

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

/sapmnt/DVQ/global/security/dvd_conn # cat kerberos.cfg
[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
}

SSL Certificates for Java

If your Hadoop cluster has full TLS encryption enabled, it’s necessary to create Java truststore 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

Server certificates of hosts, that Glue will communicate with, need to be placed in this truststore. An alternative option is to copy the complete jssecacerts truststore from any Hadoop node and place it in this path.

SAP prerequisities

Kerberos cookie encoding

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

ict/disable_cookie_urlencoding = 1

The parameter is static.

SAP RFC role and user

The Java connector uses a dedicated user in the SAP system for communication. In our reference configuration, we use the username 'hadoop'. This user needs to be created with type 'Communications Data' and with authorizations limiting his privileges to basic RFC communication.

Authorization object required is S_RFC with these settings:

• ACTVT = 16
• RFC_NAME = SYST, RFC1, SDIFRUNTIME
• RFC_TYPE = FUGR

Example of custom SAP role in PFCG transaction (Display Authorization Data):

SSL for SAP RFCs

To enable SSL communication for SAP RFCs, client certificates of target Hadoop nodes need to be added to SAP certificate list in transaction STRUST.

If HttpFS is used, client certificate of the HttpFS host is needed.

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

To import necessary certificates:

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

Following parameters need to be checked (typically it's running and set):

HTTP service and ICM parameters

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

There are two particularly important parameters affecting HTTP communication of SAP system:

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

SAP gateway access

External communication with SAP system goes through the SAP gateway. If SAP system parameter gw/acl_mode is enabled, there are two files (secinfo and reginfo) which limit the access.

In such case two programs need to have granted access either explicitly or by wildcard definition:

• AUTH_CONN
• JAVA_CONN

The program IDs are defined later in this guide within the step TCP/IP RFC creation. More information on the SAP gateway ACL topic can be found on the SAP web site:

https://help.sap.com/saphelp_nw73/helpdata/en/e2/16d0427a2440fc8bfc25e786b8e11c/content.htm

CONFIGURATION

When all prerequisites are fulfilled, further configuration is done from within the SAP system.

RFC Destinations

There are three RFC connections which need to be created via transaction SM59.

HttpFS RFC

This RFC connection is used for communication with Hadoop's HttpFS service which mediates operations in HDFS.

The name and description of the destination is optional, but it is recommended to designate its purpose with keywords 'Hadoop' and 'HttpFS'. In our example, the RFC destination also contains the Hadoop server hosting 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 (default is 14000)
• Path Prefix – this string consists of two parts
  1. /webhdfs/v1 part is mandatory
  2. /user/dvqhdp part defines Hadoop user's 'root' directory in HDFS where flat files from SAP system will be loaded

If SSL is used: It’s necessary to enable SSL and add client certificate list to be used in Logon & Security tab.

Authentication RFC

This RFC connection is part of the authentication mechanism towards any Hadoop cluster in a kerberized environment. 

The RFC setup is very basic as parameters affecting authentication are defined elsewhere. It is recommended to use the generic RFC name 'HADOOP_AUTH_CONN':

Entries explained:

• Connection Type – T for TCP/IP Connection
• Activation Type – select Registered Server Program
• Program ID – AUTH_CONN

It is important to enable the AUTH_CONN program registration in the SAP gateway (SAP gateway access).

Java RFC

Java RFC by name refers to the Java service which is used for communication with Hadoop services. Again, the setup is basic, and parameters of the Java connector are defined in separate tables:

Entries explained:

• Connection Type – T for TCP/IP Connection
• Activation Type – select Registered Server Program
• Program ID – JAVA_CONN

It is important to enable the JAVA_CONN program registration in the SAP gateway (SAP gateway access).

Java connector setup

Java connectors are configured using files that have to to be defined in SAP system.
They are configured in the following steps.

Logical file path definition

The first step is to map logical path ZHADOOP_SECURITY to the OS path where the files are stored. The actual OS path is created in the section  Datavard connector directories (/sapmnt/<SID>/global/security/dvd_conn).

Kerberos logical file definition

Before actually setting up the Hadoop storage, there are three files which are required for successful Kerberos authentication. They need to be defined as logical names to the SAP system via the FILE transaction.

When the logical path is defined, file definition follows:

ZHADOOP_KRB_KEYTAB and ZHADOOP_KRB_CONFIG refer to Kerberos keytab of <sid>hdp user and Kerberos configuration file defined in section Kerberos keytab and configuration files respectively. ZHADOOP_CDH_DRIVER refers to the custom Cloudera driver configuration file, which will be generated during the storage activation.

SSL logical file definition

If Hadoop services cluster resides in safe environment which is accessible only with SSL authentication, the following logical file needs to be defined as follows:

Drivers logical file definition

As described in JDBC Drivers section, JDBC drivers for Hadoop services connection are stored on the operating systems underlying SAP system. They also need to be defined as logical names to the SAP system via the FILE transaction.

In our example, we will be using Hive and Impala JDBC Drivers provided by Cloudera. The first step is to map logical path ZJDBC_DRIVER_PATH to the OS path where the files are stored (in our case /urs/sap/<SID>/dvd_conn/drivers/).

Example:

When the logical path is defined, a definition of driver specific folders follows:

ZJDBC_HIVE_CLOUDERA_JARS and ZJDBC_IMPALA_CLOUDERA_JARS refer to the folders in which Hive JDBC drivers and Impala JDBC drivers provided by Cloudera have been placed in section JDBC Drivers.

Java connector configuration

There are two Java connectors which Datavard Glue works with. One is used for the Kerberos authentication, the other runs SQL queries against Hadoop services. They run as two autonomous Java processes with their own log file and they are configured by setting up parameters in three configuration tables, which are imported to the SAP system as part of the Datavard Glue transports.

The ones which need to be configured manually are:
/DVD/HDP_CUS_C – contains information for connector authentication, it connects HttpFS RFC with authentication files.

The table can be maintained via transaction SM30.

Sample entry:

Entries explained:

• Destination – HttpFS RFC destination created in HttpFS RFC
• User Name – Hadoop user principal created in Hadoop user, group and HDFS directory
• Auth. Method – authentication method towards the Hadoop cluster
• Krb. Keytab – logical file definition for Kerberos keytab file
• Krb. Config – logical file definition for Kerberos configuration file
• Krb. Service RFC – authentication RFC destination created in Authentication RFC
• SSL Keystore – logical file definition for SSL keystore
• SSL Password – password for accessing SSL keystore

/DVD/JAVA_HDR – maps an RFC destination to the Datavard Java connector version.

Datavard GLUE currently uses two Java connectors, HADOOP_AUTH_CONN for the authentication process and HIVE_CDH for submitting of queries to Hive. 

Table entries can be if needed, maintained via transaction.

Table example:

/DVD/JAVA_CONFIG – stores parameters for Datavard Java connectors. The table needs to be populated with entries via transaction SE16.

Sample configuration:

Prerequisites:

Multiple entries from /DVD/JAVA_CONFIG table define location and filename where a specific file will be generated. These will be marked as (generated) in detailed explanation below.

Please note, that as a prerequisite, all the paths specified in the table must be created beforehand (only folders).

Entries explained:

• Client – ID of the SAP system source data client
• CONFIG_AS_PATH – path and configuration file specifying authentication information towards the SAP system to Java connector (generated)
• CONFIG_PATH – path and configuration file specifying program ID and SAP gateway information to Java connector (generated)
• CONN_COUNT – number of connections registered at the gateway
• JAR_PATH – path and filename of Datavard Java connector JAR-file. This jar is generated (content is part of transport) in the specified folder and filename. (to be generated)
• JAVA_EXE – path to Java binary used by <sid>adm user
• LOG_FILE – path and filename of log file related to the Java connector (generated)
• MAX_RAM_USED – the amount of memory dedicated to the Java connector
• USERNAME – RFC user defined in SAP RFC role and user
• PASSWORD – password of an RFC user (defined above) in hashed form. Use instructions from chapter "Password hash generator" to generate a hash.
• PEAK_LIMIT – maximum number of connections that can be created for a destination simultaneously
• PROG_ID – registered program ID defined in Authentication RFC (version 32) and Hive RFC (version 34)
• REP_DEST – Java connector client destination from which to obtain the repository
• WORK_THREAD_MAX – number of threads that can be used by Java Connector
• WORK_THREAD_MIN –  number of threads always kept running by Java Connector
• JAVA_START_CMD - name of the system command, which starts Java
• CHMOD_CMD - name of the system command for setting executable rights

Password hash generator

Use report /DVD/XOR_GEN for this purpose.

Storage Management setup

A generic Datavard software component: “Reuse Library” is used for the setup. The required component is “Storage Management for Hadoop”.

Datavard Storage Management facilitates transparent communication with different types of storages, which includes various types of databases and Hadoop: HDFS for flat files and Hive for structured data.

HDFS storage

In order to transparently store data, two types of Hadoop storages are requeired to be defined in Storage Management:

• HDFS storage which facilitates transfer of files to HDFS via Hadoop's 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 which connects to Impala agents to provide fast SQL execution by leveraging Impala in-memory data caching

HDFS storage creation is done via transaction:

/DVD/SM_SETUP > [Edit mode] > [New storage]

Entries explained:

• Storage ID – name of the storage
• Storage Type – choose HADOOP for HDFS
• Description – extended description of the storage for easier identification
• RFC Destination – HttpFS RFC destination defined in HttpFS RFC

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

Hive metastore storage

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

Entries explained:

• Storage ID – name of the storage
• Storage Type – choose SM_TRS_HV2 for Hive
• Description – extended description of the storage for easier identification
• Database – Hive database created in Hive database
• Hive host – Hadoop server hosting the Hive service
• Hive username – Hadoop user created in Hadoop user, group and HDFS directory
• Hive password - password for Hive user
• Impala host – Hadoop server hosting the Impala service
• Impala username – Hadoop user created in Hadoop user, group and HDFS directory
• Staging location type - Storage location for data staging area (external csv tables)
  
• Staging location url (non-default) - URL address for data staging area (e.g. Azure DataLake)
  
• HTTP RFC Destination – HttpFS RFC destination defined in HttpFS RFC
• HTTP RFC Destination (HA) - HttpFS RFC destination (High Availability)
• Java connector RFC – Hive RFC destination defined in Hive RFC
• Load Engine - Engine used for loading (writing) data, e.g. Hive or Impala
• Read Engine - Engine used for reading data, e.g. Hive or Impala
• Load driver classname - Classname of the driver used for loading (e.g. Cloudera Hive - com.cloudera.hive.jdbc41.HS2Driver)
• Load driver path - Logical name of Load driver file
• Read driver classname - Classname of the driver used for reading (e.g.Cloudera Impala - com.cloudera.impala.jdbc41.Driver)
• Read driver path - Logical name of Read driver path
• Use custom connection string checkbox - if checked, use custom connection string
• Custom connection string - standard settings are ignored, custom connection string is used instead
• Use Kerberos checkbox – checked in case the Hadoop cluster is Kerberized
• Kerberos config file path – Logical name of Kerberos configuration file defined in Kerberos logical file definition
• Hive service principal – Kerberos principal of the Hive service, must reflect the Hive host
• Impala service principal - Kerberos principal of the Impala service, must reflect the Impala host
• Kerberos keytab path – Logical name of Kerberos principal keytab file defined in Kerberos logical file definition
• Kerberos user principal - User name (kerberos principal) defined for connection to Hadoop
  
• File Type – file format in which Hive will store table data on HDFS
• Compression codec - Compression codec used for storing data on HDFS
  
• HDFS Permissions - UNIX permissions for files created on HDFS
• Use Cloudera drivers checkbox – always checked if the Hadoop cluster is Cloudera distribution
• Cloudera driver config path – Logical name of the Cloudera driver configuration file defined in Kerberos logical file definition
• Skip trash - checked if HDFS files should NOT be moved to trash after deleting them
• Impala port - Impala JDBC port
• Hive port - Hive JDBC port
• Hints for hive/impala - Hints that can be specified for JDBC connection (e.g. SYNC_DDL=TRUE)
• Open cursor logic - Select which logic will be used for reading viac cursor
• Repetition for HDFS - number of times HDFS request will be repeated in case of failure
• Use compression on transfer - checked in case compression is used for filed created on HDFS
• Compression level - level of compression (0-minimum, 9-maximum)
• Use SSL - checked if SSL authentication should be used
• SSL keystore path - logical name of the SSL keystore file
• SSL keystore password - password to keystore
• Force file cursor reader checkbox (expert setting) - cursor reader is used all the time when reading data from Hadoop
• Use extended escaping checkbox (expert setting) - extending escaping is used all the time when writing data to Hadoop

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.

Troubleshooting

In case of any issues with Java connector, Java logs can be read from an application server to determine source of issue with report /DVD/SM_HIVE_DISPLAY_JAVA_LOG.

Entries explained:

• RFC Destination - Destination of rfc with java service
• Num. of last lines to display - How many last lines of log should be displayed
• Name of file with log - Filename of file with java log. (in default destination)
• Is log archived? - Check if need to display archived log (date obligatory)
• Date of log - Date of archived log to display