You might encounter errors similar to the following when you use S3 data and do the following:

• Specify an HDFS read. The data is read with Hive after sending a message similar to the following:
  Streaming read DISABLED: Cannot access directory /bakemp/hive/test/cars

• Specify HDFS_TEMPDIR= in the LIBNAME statement. In this scenario, an error similar to the following is returned:
  ERROR: Failure accessing HDFS_TEMPDIR directory: Physical file does not exist, s3a://<bucket>/subdir 

• Use PROC HDMD FROM= to generate a .sashdmd file from a Hive table. In this scenario, s3a://<bucket> is removed from the S3 path in the input directory for the generated .sashdmd file.

• Use DBCREATE_TABLE_LOCATION= to specify an S3 path. In this scenario, the following error is received:
  ERROR: java.lang.IllegalArgumentException: Wrong FS: s3a://xxx-bff/bakemp/temp/snacks, expected:
         hdfs://ip-10-251-52-210.ec2.internal:8020

• Run the DATA step in Hadoop. In this scenario, you might see errors similar to the following: 
  NOTE: Attempting to run DATA Step in Hadoop.
  ERROR: Map/Reduce job failed.  Could not run Hadoop job.
  FATAL: Unrecoverable I/O error detected in the execution of the DATA step
  program.  Aborted during the COMPILATION phase.
  ERROR: org.apache.hadoop.mapred.InvalidInputException: Input path does not
  exist:
  hdfs://ip-nn-nnn-nn-nnn.xx5.internal:8020/bakemp/hive/test/ds1.db/d2ipd002

  or 

  20181107:11.23.53.43: 00000006:           Dispatching Embedded Process job.
  20181107:11.23.57.50: 00000006:ERROR:
  org.apache.hadoop.mapred.InvalidInputException: Input path does not exist:
  hdfs://ip-nn-nnn-nn-nn.xxx.internal:8020/bakemp/hive/test/hpsumm.db/hpds2
  20181107:11.23.57.50: 00000006:           There was failure during Embedded
  Process job dispatch.

• Use SAS^® Scoring Accelerator for Hadoop, which generates an error similar to the following:  
  hivepath: s3a://<bucket>/bakemp/hive/test
  dbpath: s3a://<bucket>/bakemp/hive/test/model.db
  ERROR: java.lang.IllegalArgumentException: Wrong FS:
  s3a://<bucket>/bakemp/hive/test/model.db, expected:
         hdfs://ip-nn-nnn-nn-nnn.xxx.new:8020
  debug_debug: s3a://<bucket>/bakemp/hive/test/model.db not present in hdfs

  When this problem was encountered, the "model" database existed in Hive and the model.db directory existed in the specified S3 location: s3a://<bucket>/bakemp/hive/test/model.db.

Click the Hot Fix tab in this note for a link to instructions about accessing and applying the software update.

Operating System and Release Information

An update for this issue is available for SAS Viya 3.4. For instructions on how to access and apply software updates, see the Updating Your SAS Viya software section in the SAS Viya 3.4 for Linux Deployment Guide at

A fix for this issue for SAS/ACCESS Interface to Hadoop 9.46 is available at:

A fix for this issue for SAS/ACCESS Interface to Hadoop 9.45 is available at: