Turns out this was a bug in the Phoenix Query Server (technically Apache Avatica): https://issues.apache.org/jira/browse/CALCITE-1458 The author of the Phoenix-Sharp (C# driver) https://github.com/Azure/hdinsight-phoenix-sharp had a similar problem to what you were saying. What happens is that a change in serialization was not made fully backwards compatible and the attribute which the phoenixdb driver was likely expected (I have not verified this) was no longer being serialized over the wire. As such, the server was sending the data back in an attribute the client was not aware existed. I have a fix I'm putting up on CALCITE-1458 shortly which should address the issue without requiring a change in the client. ... View more

Apache Hive and Apache HBase are fundamentally different systems with completely different architectures. As such, which is most efficient really depends on the application use cases. It's impossible to generically state that Hive or HBase is better/worse than the other and the fact that they both use HDFS for storing data is irrelevant. Please quantify the application requirements you have if you'd like an answer about whether Hive or HBase are better for you. ... View more

It appears that this is a hard-coded check in the client side HBase Mutation class. I would recommend that you consider why you are creating such a large rowKey in the first place. There is some reading you could also do on the subject: http://hbase.apache.org/book.html#keysize ... View more