Pre-requisites for Kerberos integration

Before you begin, please verify the following:

1. The user is created and enabled on each node in the Hadoop cluster.

   NOTE: If LDAP is enabled, the user should be created in the same realm as the cluster.

2. On the , the directory /opt/trifacta is owned by the user .

3. The user exists on each node in the Hadoop cluster.

   NOTE: The must have the same user ID and group ID on each node in the cluster. Depending on your cluster's configuration, this requirement may require an LDAP command. Configuring LDAP is beyond the scope of this document.

4. The user must be a member of any special group that is permitted to access HDFS or to run Hadoop jobs.

Configure the KDC

Steps:

1. On your KDC node, configure a Kerberos principal for the :
   1. The principal's identifier has two parts: its name and its realm. For example, the principal trifacta@HADOOPVAL.MSSVC.LOCAL has the name trifacta and the realm HADOOPVAL.MSSVC.LOCAL.
   2. Retain the name and principal for later configuration.

2. Create a keytab file for the . Command:
   

   kadmin xst -k trifacta.keytab <full principal identifier>

   where:
   <full_principal_identifier> is the principal identifier in Kerberos.
   

   On the KDC, you may have to run kadmin.local instead of kadmin. The rest of the arguments should remain the same.

   NOTE: If you're creating a keytab file in an AD environment, alternative instructions may need to be applied. See below.

3. Verify that the keytab is working. Command:
   

   klist -e -k -t trifacta.keytab

4. Copy the keytab to the in the following directory:
   /opt/trifacta/conf/trifacta.keytab
   

5. Configure the keytab file so that it is owned by the user. It should only be readable by that user.
   

   NOTE: Verify that all user principals that use the platform are also members of the group of the keytab user.

Create keytab in Active Directory environments

Some additional instructions are provided for the following environments.

For MIT Kerberos

See https://kb.iu.edu/d/aumh :

> ktutil
  ktutil:  addent -password -p username@EXAMPLE.COM -k 1 -e rc4-hmac
  Password for username@EXAMPLE.COM: [enter your password]
  ktutil:  addent -password -p username@EXAMPLE.COM -k 1 -e aes256-cts
  Password for username@EXAMPLE.COM: [enter your password]
  ktutil:  wkt username.keytab
  ktutil:  quit 

For Heimdal Kerberos

  > ktutil -k username.keytab add -p username@EXAMPLE.COM -e arcfour-hmac-md5 -V 1

If the keytab created in Heimdal does not work, you may need an aes256-cts entry. In this case, locate a machine with MIT Kerberos, and use the MIT Kerberos method instead.

Configure the for Kerberos

Locate the kerberos section, which controls Kerberos authentication.

Example configuration:

Substitute your own values in place of the example values as appropriate.

"kerberos.enabled": true,
"kerberos.principal": "trifacta",
"kerberos.kdc": "kdc.mssvc.local",
"kerberos.realm": "HADOOPVAL.MSSVC.LOCAL",
"kerberos.keytab": "/opt/trifacta/conf/trifacta.keytab"
"kerberos.principals.hive": "<UNUSED>",
"kerberos.principals.namenode": "nn/_HOST@EXAMPLE.COM"
"kerberos.principals.resourcemanager": "<YOUR_VALUE_HERE>",

Parameter	Description
enabled	To enable Kerberos authentication, set this value to true.
principal	The name part of the principal you created in the KDC
kdc	The host of the KDC
realm	Realm of the KDC
keytab	Directory in the where the Kerberos keytab file is stored
principals	List of jobtrackers and namenodes that are governed by Kerberos NOTE: kerberos.principals.hive is unused. This value must be inserted into the Hive connection definition. See Create Hive Connections. NOTE: If you don't know the values to use here, see Set principal values below. NOTE: If you don't specify principal names in the principals definition section, the default names are used: mapred/<jobtracker host>@<realm>. You should specify the principals explicitly.

At this point, you should be able to load files from HDFS and run jobs against the kerberized Hadoop cluster.

Set principal values for YARN

Check the following Hadoop config properties in yarn-site.xml :

principals.jobtracker = yarn.resourcemanager.principal
principals.namenode = dfs.namenode.kerberos.principal