On clusters where Hive is using a remote DB and may be slower, we are seeing deploy failures. This apparently is caused by a 15-second timeout in an http-client call to Explore Service while enabling explore on dataset. We have seen this on a CM cluster using a remote DB and on an HDInsight cluster (remote DB). Integration test run with several of these failures: http://builds.cask.co/browse/CDAP-ITCC5-12
We should consider lengthening, or making configurable, this timeout.