Waqar,

After enabling the ICU option manually, it's not working for my huge data until I deleted the most of the biblio records.

Then only difference was, I modified <icu_chain locale="en"> to blank <icu_chain locale=""> instead of <icu_chain locale="en_IN.UTF-8">. As I am not sure what to do with locale for other languages

You cannot not have a locale, so far as I know. Leave it as locale="en" if you don't know what locale you want. 

So,

1 - I am getting these warining? Although I have already renamed my
    default.idx located at /home/koha/koha-dev/etc/zebradb/etc (i hope this is the right location)

I'm not sure why you're talking about renaming default.idx, but you should not be. default.idx is required whether you are using ICU or not. You just have to adjust whether it is using an icuchain or charmap file.

====================
REINDEXING zebra
====================
12:22:45-10/01 zebraidx(2057) [warn] Unknown register type: 0
12:22:45-10/01 zebraidx(2057) [warn] Unknown register type: n
12:22:45-10/01 zebraidx(2057) [warn] Unknown register type: y
12:22:45-10/01 zebraidx(2057) [warn] Unknown register type: d

2 - Right after that I have this warning

    [warn] previous transaction didn't reach commit

That will be because either A) there was a very corrupted record or (more likely) B) the lack of a default.idx file is giving Zebra conniptions.
 
3 - I deleted the records upto just 9000. Now zebra started to index other languages. Why it is working only on this small set of data? Do i need to define any escape sequence somewhere for some special characters.

I had problems using ICU indexing for datasets larger than ~400k. I believe the problem was a corrupted record, but with 450k records I couldn't be bothered to track down which record it was.

Regards,
Jared

--
Jared Camins-Esakov
Bibliographer, C & P Bibliography Services, LLC
(phone) +1 (917) 727-3445
(e-mail) jcamins@cpbibliography.com
(web) http://www.cpbibliography.com/