FYI.
As a result, zeppelin doesn't need to config about kerberos.
Only system and spark config needs.
Thanks.

---------- Forwarded message ----------
From: Hyung Sung Shim <[email protected]>
Date: 2015-12-04 11:57 GMT+09:00
Subject: Working zeppelin with Spark on YARN in the kerberos evnvironment.
To: "[email protected]" <[email protected]>


Hello.

I've been testing about spark on yarn in the kerberos environments as
zeppelin backend .
Let me share you.

*Infrastructure.*
I set up KDC(kerberos distribution center) and Hadoop system(hdfs, yarn
resource manager and nodemanager), spark and zeppelin.
[image: 본문 이미지 3]

*Setting and Configuration.*
1. Zeppelin system(OS) needs to install kerberos client modules and
configuration"krb5.conf" because aware of KDC.

2. Set SPARK_HOME in zeppelin-env.sh (to use spark-submit).

3. Add spark configuration(~/conf/spark-defaults.conf) below two properties.

>
> spark.yarn.principal
> spark.yarn.keytab


4. Play with zeppelin.

That's it.

If you have any questions feel free to mail me.

Thanks.

Reply via email to