now that all the concurrency bugs are "solved" with hive server 2, I see no reason why the lighter thrift action I made is not the default.
http://www.edwardcapriolo.com/roller/edwardcapriolo/entry/hive_and_oozie_featuring_hive On Mon, Mar 4, 2013 at 1:40 PM, Paul Chavez <[email protected] > wrote: > Hello, > > I've had some coordinators running since the first of this year that run a > nightly sqoop and then some hive scripts afterward. Last week our cluster > was upgraded to CDH 4.2 and HDFS HA, since then the actions fail on the > first data load step. We also have always had a MySQL metastore database > for hive. > > I'm pretty sure the issue lies with accessing the metastore, since the > first LOAD statement actually moves the files to the new location but then > fails with a qlMove exeption and the partition never appears in the table > list. > > Previously I was attaching 4 javax.jod.option properties (ConnectionURL, > ConnectionDriveName, ConnectionUserName, ConnectionUserPassword) to the > hive action itself to get this to work. So far I have verified that the > MySQL connector JAR is still in the oozie share lib as it was before. If I > remove the 4 jdo properties I get a 'table not found' error. I recently > updated the connectionURL slightly to the new format in hive-site.xml > (appending some query params for encoding) and now I get a 'Exception while > processing '<table_name>'. > > Previously hive was managed by 'us', now it's integrated with CM 4.5. We > also moved from hive-server to hive-server2. I know a lot of things changed > at once, so hoping for some guidance as I'm a bit lost as to what needs to > be fixed here. > > thank you, > Paul Chavez > >
