The script spent 20.91 of 21.19 minutes on the import process.
I guess we can rule out delete and vacuum as the source of problem.
Expecting more questions like yours, I did a little more research. The
import is all INSERT's except for two selects. I realised I could rewrite
a portion of the code to remove one of those SELECT's. Guess What no
effect on run-time. Isn't that curious? So I pondered that for a while
and thought that if removeing one SELECT didn't make any difference, then
making changes to the second SELECT (which was on same table) wsn't likely
to change anything either. If that's true then the problem was only
INSERT's. A little more pondering and I remember that I looked a little
closer at startup parms during the PG upgrade. I had removed -F option.
Put it back in and runtime dropped to 10 minutes again - problem solved!
Frank
Brent R. Matzelle <bmatzelle@yahoo.com> wrote:
>--- Frank Bax <fbax@sympatico.ca> wrote:
>> The script just deletes all rows in tables, imports
>> data as seperate inserts for each row, then does vacuum
>> analyse on the tables. I've read that delete/rebuild
>> of indexes will likely improve performance of this
>> script, but the runtime itself is not really a concern
>> here, just the change in runtime between releases. I'm
>> curious what changed between releases to cause the
>> slower processing?
>
>Your script performs 3 separate actions. Do you know which one
>(delete, import, vacuum analyze) is slower when using the new
>version of PG? That will help pinpoint the problem so that we
>can more easily solve the problem.
>
>Brent