I just successfully completed upgrading our cluster to CDH 5.6.0. (starving developer's version) . Running Hive on Spark Yarn. I am happy to announce that since all the installations I have done from CDH 4.1.2 thru now, today was the only install with ZERO issues. I am very excited !
I would like to thank Cloudera very (very) much for creating and continuously improving their kickass best-in-class hadoop distribution and a superb installation and management console. And lucid documenattion ! Kudos for that too.
And special thanks to the quick, patient and helpful responses from Cloudera developers to many (many) of the questions I have asked on the group these past 4 years . I never once felt the responses were slow and I have always worked on the starving developer version.
On this 10th anniversary of hadoop, a massive applause to Doug Cutting ! Go Cloudera !
warmly and appreciatively,
p.s. : When did this grow to be an redbook size document - 794 pages ! Wow ! :-)
We thoroughly appreciate your kind words and for taking the time to give us this most valued feedback! I have shared this with our engineering and quality teams and I think you made their day! We invest heavily in quality so your affirmation that it's helping is very rewarding for our team members who work tirelessly on this.
I'm also glad you're finding the community to be helpful in your big data journey.
You are most welcome Clint.
I wish there is some way to migrate all of the google groups cloudera,cdh discussions into this community because that is still my reference :-) where I go to see for example - what Romain said about that Hue setting of what Robert (Kanter) said about a hive server question I had posted ? :-)
At this point we are heavily relying on Hive(on Spark) , Impala and HBase to create multiple sliced and diced versions of input data sets which we feed into our modeling processes. We have REST services based on Hbase behind that one can now get input datasets on demand. Its pretty cool. I love it. Machine Learning is still 80% data wrangling. Just as Clapton wont sound good with rusted guitar strings or a detuned guitar, machine learning wont work with bad data !
thanks once again as always,