Known Issues and Restrictions

Sqoop importing data into Hive fails when the entire cluster is configured to use Tez.
This is because of Sqoop's incompatibility with Tez.

Workaround: Do not configure the entire cluster to use Tez.

Percentage sampling is not supported in org.apache.hadoop.hive.ql.io.HiveInputFormat.

Hive uses org.apache.hadoop.hive.ql.io.HiveInputFormat by default and so queries like 'SELECT * FROM tablename TABLESAMPLE(20 percent);' will not work for Hive on Tez.

Workaround: Instead of org.apache.hadoop.hive.ql.io.HiveInputFormat, use org.apache.hadoop.hive.ql.io.CombineHiveInputFormat.

To change input format, do one of the following:

  • Set hive.tez.input.format in hive shell. For example:
    hive> set hive.tez.input.format=org.apache.hadoop.hive.ql.io.CombineHiveInputFormat;
  • Add org.apache.hadoop.hive.ql.io.CombineHiveInputFormat to hive-site.xml file. For example:
    <property>
      <name>hive.tez.input.format</name>
      <value>org.apache.hadoop.hive.ql.io.CombineHiveInputFormat</value>
    </property>
Hive on Tez does not work well with Sequence Files Schema changes
TEZ-2741
Limitations with common joins
HIVE-11693: The CommonMergeJoinOperator only sets big table position when it has inputs for big table. If the input is empty, the method is not called.
HiveServer2 on Tez doesn't support concurrent queries within one session
HIVE-9223: When multiple queries are submitted in the same HS2 session concurrently, some queries fail with an error.
Tez upgrade issues
  • No support for preserving configuration from EEP-5.0.0 and EEP-4.1.1 (ECO-1803) to EEP-6.0.0(1808) or EEP-5.0.1(1808) on Ubuntu.
  • No support for preserving Tomcat configuration from previous EEPs to EEP-6.0.0 (1808).
  • You should manually stop the Tomcat service and delete the tomcat folder as a precondition if you are updating or upgrading Tez from the following EEPs:

    EEP-4.0.0

    EEP-4.1.0

Tez shuffle SSL encryption issue
During a shuffle phase, the javax.net.ssl.SSLException error could occur on a multi-node cluster due to insufficient Tez shuffle SSL encryption configuration, see Tez Shuffle for a solution.