Mapreduce with third party API - hadoop

Question

I am using some third party APIs in my Map-Reduce code. I tried few things but didn't worked out for me. (I am using Cloudera 5.9 version)
1) I tried with fat jar approach (building jar with all dependencies included), and my code worked for me, but this is not an good way to use (my jar size in very heavy).
so i thought of separating the third party jars and share them using distributed cache.
2) I tried using the some options like below.
i) I used hadoop jar <jar_file_name> <main_class_name> <arguments> -libjars <list of jar files> command. --> Didn't worked and getting ClassNotFoundException
i.a) I Kept the third party jars in local folder and used the path in -libjars option. --> Didn't worked and getting ClassNotFoundException
i.b) I Kept the third party jars in HDFS and used the path in -libjars option. --> Didn't worked and getting ClassNotFoundException
ii) I Updated the code to use DistributedCache.addFileToClassPath() --> Didn't worked and getting ClassNotFoundException
iii) I Updated the code to use job.addCacheFile() --> Didn't worked and getting ClassNotFoundException
iv) I Updated the /etc/hadoop/conf/hadoop-env.sh and added a new line export HADOOP_CLASSPATH=$HADOOP_CLASSPATH:/home/mani/test/lib/* and then restarted the cluster and but still didn't worked.
v) I tried running the command like hadoop jar <jar_file_name> <main_class_name> <arguments> -D mapred.child.env="LD_LIBRARY_PATH=path/to/my/lib/*" --> Didn't worked and getting same exception.
vi) I Also tried -Dyarn.application.classpath but still same issue.
I have gone through some forums and Cloudera blog posts, and some other websites. Everyone telling the same story but i am not able to get the output even after following their posts correctly, am i missing something here.
Can someone please help me to find out the solution.
Thanks,
Manindar.