Other Miscellaneous Changes in Impala

Review the changes to Impala syntax or service that might affect Impala after migrating your workload from CDH version 5.13-5.16 or CDH version 6.1 or later to CDP Private Cloud Base or CDP Public Cloud. The version of Impala you used in CDH version 5.11 - 5.16 or 6.1 or later changes to Impala 3.4 in CDP Private Cloud Base.

Decimal V2 Default

In CDP, Impala uses DECIMAL V2 by default.

To continue using the first version of the DECIMAL type for the backward compatibility of your queries, set the DECIMAL_V2 query option to FALSE:

Column Aliases Substitution

To conform to the SQL standard, Impala no longer performs alias substitution in the subexpressions of GROUP BY, HAVING, and ORDER BY.

The example below references to the actual column sum(ss_quantity) in the ORDER BY clause instead of the alias Total_Quantity_Purchased and also references to the actual column ss_item_sk in the GROUP BY clause instead of the alias Item because aliases are no longer supported in the subexpressions.

    ss_item_sk as Item,
    count(ss_item_sk) as Times_Purchased,
    sum(ss_quantity) as Total_Quantity_Purchased
    from store_sales
    group by ss_item_sk
    order by sum(ss_quantity) desc
    limit 5;
    | item  | times_purchased | total_quantity_purchased |
    | 9325  | 372             | 19072                    |
    | 4279  | 357             | 18501                    |
    | 7507  | 371             | 18475                    |
    | 5953  | 369             | 18451                    |
    | 16753 | 375             | 18446                    |


The PARQUET_ARRAY_RESOLUTION query option controls the behavior of the indexed-based resolution for nested arrays in Parquet. In Parquet, you can represent an array using a 2-level or 3-level representation. The default value for the PARQUET_ARRAY_RESOLUTION is THREE_LEVEL to match the Parquet standard 3-level encoding. See Parquet_Array_Resolution Query Option for more information.

Clustered Hint Default

The clustered hint is enabled by default, which adds a local sort by the partitioning columns in HDFS and Kudu tables to a query plan. The noclustered hint, which prevents clustering in tables having ordering columns, is ignored with a warning.

Query Options Removed

The following query options have been removed:

Shell Option refresh_after_connect

The refresh_after_connect option for starting the Impala Shell is removed.


The EXTRACT and DATE_PART functions changed in the following way:
  • The output type of the EXTRACT and DATE_PART functions changed to BIGINT.
  • Extracting the millisecond part from a TIMESTAMP returns the seconds component and the milliseconds component. For example, EXTRACT (CAST('2006-05-12 18:27:28.123456789' AS TIMESTAMP), 'MILLISECOND') returns 28123.

Port for SHUTDOWN Command

If you upgraded from CDH 6.1 or later and specified a port as a part of the SHUTDOWN command, change the port number parameter to use the Kudu7: RPC (KRPC) port for communication between the Impala brokers.

Change in Client Connection Timeout

The default behavior of client connection timeout changes after upgrading.

In CDH 6.2 and lower, the client waited indefinitely to open the new session if the maximum number of threads specified by --fe_service_threads has been allocated.

After upgrading, the server requires a new startup flag, --accepted_client_cnxn_timeout to control treatment of new connection requests. The configured number of server threads is insufficient for the workload.

If --accepted_client_cnxn_timeout > 0, new connection requests are rejected after the specified timeout.

If --accepted_client_cnxn_timeout=0, client waits indefinitely to connect to Impala. This sets pre-upgrade behavior.

The default timeout is 5 minutes.

Interoperability between Hive and Impala

Impala supports a number of file formats used in Apache Hadoop. It can also load and query data files produced by other Hadoop components such as Hive. After upgrading from any CDH 5.x version to CDP Private Cloud Base 7.1, if you create an RC file in Hive using the default LazyBinaryColumnarSerDe, Impala can not read the RC file. However you can set the configuration option of hive.default.rcfile.serde to ColumnarSerDe to maintain the interoperability between hive and impala.

Improvements in Metadata

After upgrading from CDH to CDP, the on-demand use_local_catalog mode is set to True by default on all the Impala coordinators so that the Impala coordinators pull metadata from catalogd and cache it locally. This reduces memory footprint on coordinators and automates the cache eviction.

In CDP, the catalog_topic_mode is set to minimal by default to enable on-demand metadata for all coordinators.

Recompute the Statistics

After migrating the workload from any CDH 5.x version to CDP Private Cloud Base 7.1, recompute the statistics for Impala. Even though CDH 5.x statistics are available after the upgrade, the queries do not benefit from the new features until the statistics are recomputed.

Mitigating Excess Network Traffic

The catalog metadata can become large and lead to excessive network traffic due to dissemination through the statestore. The --compact_catalog_topic flag was introduced to mitigate this issue by compressing the catalog topic entries to reduce their serialized size. This saves network bandwidth at the cost of a small quantity of CPU time. This flag is enabled by default.