Please never disable that check. Checkpoints are very essential for the HDFS operation, and you do not want to be in a position with checkpoints failing for a technical reason and you never getting notified on that.   Instead, look at your Standby or Secondary NN to figure out what the error is, and/or seek help with that identified information. ... View more

Could you share the full exception trace? It appears your fsimage, or one of your edit logs, has somehow gotten some corrupt data in it. If its the edit log, you can perhaps attempt to skip a few entries in the 'hdfs namenode -recover' startup mode; but if its the fsimage that is corrupt, then we'll need to rollback to an older copy and replay edit logs on top. The full exception trace can help tell which file is corrupt. ... View more

CPUs are considered equally, if the request seeks that. ... View more

Yes, do you not see it working? You'll need to pass the XML property via the workflow.xml under the action's configuration section. ... View more

Thanks, that'd explain your transition. What application is this? Is it an MR2 application, Spark app, or something custom? ... View more

Glad to hear; Please consider marking the topic as resolved so others with similar issues can find it easily. ... View more

Unlike your spark shell command, Oozie does not invoke/use scripts that setup local classpaths for its actions (as it needs to use distributed-caches for this). Take a look at how the ShareLib works, and how you can override them for your action to include a system one http://archive.cloudera.com/cdh5/cdh/5/oozie/WorkflowFunctionalSpec.html#a17_HDFS_Share_Libraries_for_Workflow_Applications_since_Oozie_2.3. In your case, if you use the java action, you can make it include the "hive" share-lib, and that will include all Hive jars into the distributed cache classpath. ... View more