"John Watts" <jwatts@promotion-update.com> writes:
> Anyone?
I'm still suspicious that you're not really re-ANALYZE'ing the relevant
tables, because there are some discrepancies in the row count estimates
that seem hard to explain otherwise, eg here:
-> Index Scan using tblcompanyindidnumber on tblcompany (cost=0.00..8.40 rows=1 width=206) (actual
time=0.003..0.003rows=0 loops=3445)
Index Cond: (tblappliccomp.companyid = tblcompany.idnumber)
vs
-> Bitmap Heap Scan on tblcompany (cost=13.07..1774.92 rows=620 width=185) (actual time=0.013..0.013
rows=0loops=3445)
Recheck Cond: (tblappliccomp.companyid = tblcompany.idnumber)
-> Bitmap Index Scan on tblcompanyindidnumber (cost=0.00..12.91 rows=620 width=0) (actual
time=0.011..0.011rows=0 loops=3445)
Index Cond: (tblappliccomp.companyid = tblcompany.idnumber)
That might be caused by missing stats for either tblcompany or
tblappliccomp. Or perhaps the problem is much different values of
default_statistics_target?
Also, I've got to say that this does not represent good practice:
> server_version | 8.3.0
You're missing eighteen minor-release updates on that server. We don't
do minor releases just to keep ourselves amused; there are a lot of
rather significant bug fixes that you're missing, possibly including
some that affect this issue.
regards, tom lane
