|
Archives of the TeradataForumMessage Posted: Mon, 01 Nov 2004 @ 10:31:52 GMT
Was the upgrade done for V2R5.1? If so, then do you have a SYSLIB database (even if it is empty)? If SYSLIB does not exist and it is 5.1 then this could be your problem. We had this problem when we first upgraded causing the tech alert copied below. I'm curious this if this is your same problem since the generic change controls were suppose to add the addition of the SYSLIB database as part of the upgrade. Let me know Kiki Sanchez ----------------------------------------------------- TA#: NTA 817 Product: Parser_Software / Condition/Symptom: With V2R5.1, sites have experienced increased parsing overhead and AMP skew. This problem is particularly evident at sites using TPUMP. Probable Cause: With V2R5.1, Teradata introduced User Defined Functions (UDFs). Support of this feature requires that we check for the existence of a UDF during the parsing phase. Usually this added overhead is insignificant relative to total processing time but, under certain circumstances, this overhead can be significant. Configuration Data: V2R5.1.0 and above. Solution: DR 88759 has been opened to address this issue. As a workaround it is recommended that all V2R5.1 sites create a database "SYSLIB". SYSLIB is a recommended, but optional, function repository database. It is not installed as part of a Teradata Database installation or upgrade. By creating this database, the overhead associated with a UDF check is significantly reduced. In addition, it allows us to cache the lookup operations which eliminates the performance impact of repeated lookups that we can incur with TPUMP . If your site is not using UDFs, SYSLIB can be created with a nominal amount of PERM space. It is not necessary to grant privileges on this database but users must have the EXECUTE FUNCTION privilege on any UDF they run from SYSLIB. References to Procedures: Additional Information/Comments: If ResUsage is enabled, the following query can be used to identify candidates for this occurrence: SELECT TheDate, NodeId,VPRId, SUM(netPtPReads) (format 'zzz,zzz,z99') FROM DBC.ResUsageSVPR WHERE TheDate > DATE-10 AND VPRId < 16000 GROUP BY 1,2,3 ORDER BY 1,4 DESC; If a single AMP consistently shows a point-to-point read count which is an order of magnitude above other AMPs, this is likely caused by repetitive UDF validation checks. This same scenario can sometimes be identified with PMON --> View --> Resource or Resource Button by comparing the Net Reads values for each AMP.
| ||||||||||||||||||||||||||||||||||||||||||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||
Copyright 2016 - All Rights Reserved | ||||||||||||||||||||||||||||||||||||||||||||||||
Last Modified: 15 Jun 2023 | ||||||||||||||||||||||||||||||||||||||||||||||||