Sep 5, 2010 3:07 PM
Not enough memory to partition hash join
-
Like (0)
Has anyone seen or know how to resolve this error: "Not enough memory to partition hash join". The query that is producing this error runs everyday and has been fine. While the tables have grown the dataset that is being worked on is pretty constant (a rolling 30 day's worth). There has been a disk failure (and completed regen of that disk to a spare) between the last successful time that this query has executed successfully and this error.
Hi,
Did you find any answer to your problem? We are facing quite similar problem. The error is: "ERROR: DHJ: not enough memory for joining". The query shouldn't be that bad..
Mika,
I was told by Netezza that a fix was made to resolve this issue in 5.0.10 (see SWS-52386 in 5.0.10 upgrade release notes). However, prior to upgrading our system I rewrote the query. The query was joining two large tables, Table A and Table B. I broke the query into a 2-step process, by first creating a new table (Table C) from Table A and applying some of the filters from the where clause of my original qurey and distributing the new table using the same distribution key as in Table B. This reduced the size and since the distribution column was my join key it offered co-located join. (The reason I didn't just simply distribute on the join column in Table A to begin with, is that this particular column does cause some data skew, and is not used in any other queries).
We have upgraded, but if I were to do a similar query again without doing the 2-step process I get the following error: Not enough sort/merge buffers.

