I am using visual recipes on HDFS partionned data. This morning we installed Impala (on a cluster cloudera and we set up the comunicationwith dataiku). Indeed we wanted to use the in-database engine to produce charts of all the available datasets (around 60GB). From now, we can't manage anymore the recipes.
1. Making sure that the various command lines (hdfs, hive, spark-shell…) work correctly from the node where Dataiku is installed. If not, it is a Hadoop configuration problem, so you should ask your Hadoop administrator or Cloudera support.
2. If it works on the command line but not in Dataiku DSS, perform a DSS diagnostic (in Administration > Maintenance) and send the result to Dataiku support http://support.dataiku.com/ along with a precise description of what doesn't work.