On Friday 24 October 2008, "Grzegorz Jaśkiewicz" <[EMAIL PROTECTED]> wrote:
> with two sata discs in software raid 1 on linux. And it seems to spend
> loads of time (40-60% sometimes) on waits. I guess this is due to lack of
> >aio support in postgresql,
No, it's due to the fact that hard disks
that index did the job, also reindexing, and getting rid of two other not
quite often used indices helped a lot. Now, this whole machine is fairly
simple two way p4, with two sata discs in software raid 1 on linux. And it
seems to spend loads of time (40-60% sometimes) on waits. I guess this is
du
> Can postgres use combined indicies for queries that would only require part of
> it ?
Even if not, if there is at least one index that reduces the potential
matches to a small set, then scanning those rows against the other criteria
won't take so long. (Assuming good stats and PG choosing a goo
On Thu, Oct 23, 2008 at 2:47 PM, Tom Lane <[EMAIL PROTECTED]> wrote:
>
>
> I suppose the problem is that rows with processed = false are very few
> in the upper range of data. If so, and if you really need this to go
> fast, a partial index might be worth its overhead:
>create index foo o
"=?UTF-8?Q?Grzegorz_Ja=C5=9Bkiewicz?=" <[EMAIL PROTECTED]> writes:
> I have a simple query over a fairly simple query here, that scans for max
> date in a table that's fairly hudge (300M rows). there's index on that field
> that's being used, but for whatever reason, it takes ages. Ideas ?
> sele
if reindex will help (which I will run overnight, and will let you know
Tomorrow) - I find it quite worrying, cos it is 8.3, and I was hoping - with
HOT in place, and all these features - that reindexing of that table
wouldn't be needed. it is 'only' 375M rows now, but I can definetively feel
pain
Hello
try VACUUM and REINDEX
regards
Pavel Stehule
2008/10/23 Grzegorz Jaśkiewicz <[EMAIL PROTECTED]>:
> hey folks
>
> I have a simple query over a fairly simple query here, that scans for max
> date in a table that's fairly hudge (300M rows). there's index on that field
> that's being used, but
