Tom Lane wrote:
> Stefan Kaltenbrunner <stefan@kaltenbrunner.cc> writes:
>> Tom Lane wrote:
>>> Apparently we've made the planner a bit too optimistic about the savings
>>> that can be expected from repeated indexscans occurring on the inside of
>>> a join.
>
>> effective_cache_size was set to 10GB(my fault for copying over the conf
>> from a 16GB box) during the run - lowering it just a few megabytes(!) or
>> to a more realistic 6GB results in the following MUCH better plan:
>> http://www.kaltenbrunner.cc/files/dbt3_explain_analyze2.txt
>
> Interesting. It used to be that effective_cache_size wasn't all that
> critical... what I think this report is showing is that with the 8.2
> changes to try to account for caching effects in repeated indexscans,
> we've turned that into a pretty significant parameter.
yes I'm a bit worried about that too - it has been a bit of
"conventional wisdom" that setting effective_cache_size optimistic will
never hurt and that it encourages postgresql to sometimes get a better
plan by favouring index-scans.
>
> It'd be nice not to have to depend on the DBA to give us a good number
> for this setting. But I don't know of any portable ways to find out
> how much RAM is in the box, let alone what fraction of it we should
> assume is available per-query.
well there are really a number of things the dba would better give
accurate information to the database - though in that case we might go
from "too much won't hurt" to "too much will hurt" ...
Stefan