Description
When S3 credentials are included in hive-site.xml, then MR jobs that need to read
data from S3 cannot use them because S3 values are stripped from the Job configuration
object before submitting the MR job.
@Override public void initialize(HiveConf conf, QueryPlan queryPlan, DriverContext driverContext) { ... conf.stripHiddenConfigurations(job); this.jobExecHelper = new HadoopJobExecHelper(job, console, this, this); }
A nice to have (available on hadoop 2.9.0) is an MR mapreduce.job.redacted-properties that can be used to hide this list on the MR side (such as history server UI) to allow MR run the job without issues.
UPDATE:
Change the call to stripHiddenConfigurations() in ql/exec/tez/DagUtils.createConfiguration(), because this is currently broken for running hive-blobstore suite against Tez
Attachments
Attachments
Issue Links
- is related to
-
HIVE-14588 Add S3 credentials to the hidden configuration variable supported on HIVE-14207
- Resolved