With SAS/ACCESS^® Interface to Hadoop, a standard part of each query includes a SELECT * FROM TABLE statement to retrieve table metadata. This query is generally very fast and has no significant overhead. However, if the hive table is large and the hive property hive.fetch.task.conversion.threshold setting is exceeded, it can cause the launch of a large MapReduce job, which would be quite costly.

To circumvent this issue:

• Add this to the SAS LIBNAME statement: properties="hive.fetch.task.conversion.threshold=-1;hive.fetch.task.conversion=more"
• If you prefer, set the hive.fetch.task.conversion property to minimal. Setting it prevents the statement from picking up a value of NONE, which would result in a MapReduce being run for all queries.

Operating System and Release Information

A fix for this issue for SAS/ACCESS Interface to Hadoop 9.43 is available at: