 |
 |
Archives of the TeradataForum
Message Posted: Mon, 23 Aug 2010 @ 21:56:12 GMT
| Subj: | | Re: SysMaxRow size? |
| |
| From: | | John_Wight |
DWellman replied:
| | Actually it might well be the "union all", I've had that before (and it was a pain to sort out). | |
| | Read through the plan VERY carefully and see if you're ending up with a spool file containing the query text from BOTH dbqlogtbl and
dbqlsqltbl (or whatever the actual names are). | |
| | Also, remember that although the query text column is only 31KB long, it is a UNICODE column (certainly in the DBC tables) so the byte
count for this alone can be up to 62KB. The worst thing that I found was that this is effectively a 'run time' error, meaning that whether or not
you get the error depends on the data. With some data it works, with other data it won't. | |
It has to be the UNION ALL - I agree. Data from DBQLLog and DBQLSql tables are present (and should be) but not the query text part - only the
DBQLSql text. Had to look and see if they were 'mixed' as one in the log table is 10000 bytes - but no, both come from the DBQLSql table. Good
point though to look at.
Also, good point about DBC dqbl -vs- PMCP dbql data type definitions. The SQL test in DBC is UNICODE (as you say) but the one in the history
PMCP table is LATIN!!! However, I can't think of any UNICODE data being saved in that column - its' all SQL we write for ETL that is 'snapped' by
the system periodically. Also, as we are a US company only and don't have UNICODE data, it's unlikely that any could get in there.
I ended up creating a MACRO with the two same queries (no UNION) with two result sets - that works great - only I get two result sets instead
of one!! But we can work with that.
Again, thanks for your time - much appreciated.
JK
| |
