Notifications - Aster Execution Engine

Teradata Aster® Execution Engine for Aster-on-Hadoop OnlyRelease Notes

Product
Aster Execution Engine
Release Number
7.00.02
Published
July 2017
Language
English (United States)
Last Update
2018-04-17
dita:mapPath
gay1497034939562.ditamap
dita:ditavalPath
External-only.ditaval
dita:id
ltx1496451827304
Product Category
Software
  • Limited Temp Space

    Because Hadoop nodes tend to have small temp partitions, you may encounter issues where temp space is completely consumed. For example, Aster Analytic function FindNamedEntity() outputs excessively to temp space when reading large amounts of data. This triggers a system shutdown and a restart.

  • ncluster_loader
    Aster Execution Engine does not support underlying features used by some optional functionality of ncluster_loader. Specifically, Aster Execution Engine has no loader node and does not allow parent/child partition via inheritance. By extension, Aster Execution Engine also does not allow custom error logging tables, since custom error logging tables are required to be created using parent/child partition via inheritance. The affected ncluster_loader optional flags are --el-table, --el-errfile, --el-schema, --auto-partitioning, --loader and --force-loader. Teradata recommends theses guidelines when using these optional flags withncluster_loader:
    • Do not use the --el-schema command line flag, because its default value is the only allowed value.
    • Unless you want to use the command line flag --el-errfile, do not use the --el-table command line flag, since its default value is the only allowed value.
    • To use the command line flag --el-errfile, you must set the command line argument --el-table or the corresponding flag to either nc_errortable_part or nc_errortable_repl, depending on whether the target table is distributed or replicated, respectively.
    • Do not use the command line argument --auto-partition, since parent/child partition via inheritance is not allowed.
    • Because Aster Execution Engine does not have a loader node, do not use the command line arguments --force-loader and --loader. Instead, use the --hostname command line argument to direct ncluster_loader to the queen node.

    These command line flags have corresponding map file flags, and these guideline recommendations apply to the map file flags as well.

  • Replication

    To check the replication factor after a cluster has been activated, use the Aster Management Console or issue an ncli sysman showrf command. See Teradata Aster® Execution Engine Aster Instance User Guide.

  • Teradata Aster AppCenter
    • AppCenter service does not start automatically after rebooting an HDP machine.

      This behavior does not occur on CDH machines.

      To work around this issue, you must manually start AppCenter.

    • When an app is created using the logic shown below, visualizations are sometimes not generated, especially after the app is initially installed.
         INSERT INTO app_center_visualizations (json)
         SELECT json FROM Visualizer (
         ON “npath_table” PARTITION BY 1
         AserFunction(‘npath’)
         Title(‘hc’)
         VizType(‘hierarchical_clustering:chord’)
         );

      To work around this issue, use hierarchical_clustering:tree instead of hierarchical_clustering:chord.

  • Teradata Aster Analytics

    The IPGeo function can use the conversion file maxmind1.2.8.jar, which is part of the Aster Analytics package. If you specify this file in the Converter argument, you must specify its full name. For example:

    SELECT * FROM IPGeo (
      ON ipgeo_1
      Converter (maxmind1.2.8.jar,
        'com.asterdata.sqlmr.analytics.location.ipgeo.maxmind1.2.8.jar')
      Accumulate ('id', 'ip')
    )

    The default conversion file, GeoLiteCity.dat, is also installed as part of the Aster Analytics package.

  • Teradata Aster R
    • An additional step is necessary to set up Teradata Aster R on a cluster for AX7.00.01.

      The instructions can be found in the Teradata Aster R Orange Book, located at: https://access.teradata.com.

      Add ODBCSYSINI=/root to the /opt/aster/third-party/R/usr/lib64/R/etc/Renviron file after you update the /opt/aster/third-party/R/usr/lib64/R/etc/ldpaths file (see section B.4.3 in the Teradata Aster R Orange Book).

    • Installed R scripts in Aster are removed on a soft restart operation. The ncli database ddl_checkpoint_all command will not restore the installed R scripts. You must reinstall scripts and tables.
    • The ta.create function supports creation of dimension, fact, or analytic tables. With the AX 7.00.01 release, tables created are always analytic regardless of the tableType argument value.
    • Since only a Database Administrator can create a schema in AX 7.00.01, by default, “public” schema is used as a temporary schema for the given session. If a server restart occurs, the temporary schema (other than public) will be deleted if it is not persisted using the ncli database ddl_checkpoint_all command.

      A db_superuser can create different temporary schemas and allocate them to different users by giving them full access on the schema. The temporary schemas can be persisted by using the ncli database ddl_checkpoint_all command.

      To create a new connection, perform a ta.connect from the R client after the server restarts.

    • You cannot use user defined tables for error logging in ta.load.csv and ta.push functions. Only the system error logging table can be used for elSchema = "" and elTable = "" arguments.