1. You need issue invalidate metadata to your entry impalad daemon (which is used by bi tool or hue to connect) after adding new partition to partitioned table. 2. Impala does prepare aggregates very fast and it can insert into table. Its time saving solution. Hive is too slow. 22.10.2013 7:29 пользователь "Jarek Jarcec Cecho" <[email protected]> написал:
> Hi Lars, > would you be open to share the use case? I was under impression that > impala is more for a real-time queries that are executed either by users or > BI tools rather then a part of larger background pipelines. > > Jarcec > > On Mon, Oct 21, 2013 at 11:49:06AM +0200, Lars Francke wrote: > > Hi, > > > > I was looking for a Impala action in Oozie but couldn't find any > > discussion about this at all. > > > > I'm thinking about opening a JIRA but I'm a) not sure if Oozie would > > accept such an action or if it should live in a third-party (e.g. > > Cloudera) extension and b) if it might be generalized into a JDBC > > action where I can provide a JDBC driver, connection details and a SQL > > script to execute. To me at least that'd have been helpful in the > > past. > > > > I'm looking for opinions & guidance on whether either option would > > make a good JIRA for the Oozie project. > > > > Thanks! > > > > Cheers, > > Lars >
