In T212259#5288214 we identified a way to make hive actions work in a Hadoop kerberized cluster. As preliminary step we should, in my opinion, slowly move all refinery actions to hive2 ones (without credentials for the moment) and make sure that they work as expected. This would surely reduce the number of errors that we could encounter when Kerberos will be enabled. Ideally this task is something that people can do while on ops duty (so moving some jobs each week as part of the weekly train deployment).
These are the available schemas for hive2 actions:
I tested 0.1 but surely 0.2 is fine as well (but we'll need to verify that). The idea is, for each hive action, to do the following:
- replace hive with hive2 (no cred field needed for now since we don't have any kerberos credential):
- <action name="add_partition"> - <hive xmlns="uri:oozie:hive-action:0.2"> + <action name="add_partition"> + <hive2 xmlns="uri:oozie:hive2-action:0.1">
- add jdbc parameter
The latter should be added via variable and configured in each .properties file.