Re: update functions locking tables

30 Aug 2005 10:35:31 -0400, Greg Stark <gsstark(at)mit(dot)edu>:
>
> Well a regular vacuum will mark the free space for reuse. If you insert or
> update any records the new ones will go into those spots. Make sure you set
> the fsm_* parameters high enough to cover all the updates and inserts for the
> entire day (or repeat the vacuum periodically even if there are no deletes or
> updates going on to create more free space).

I will check those fsm_* parameters.

>
> You should realize that what's going on here is that the old records are still
> in your table, marked as deleted. So any sequential scan will take twice as
> long as otherwise. I think even index scans could take twice as long too
> depending on the distribution of values.
>
> I'm not saying that's untenable. If all your queries are fast enough then
> you're set and it's just a cost of having no downtime.
>
> --
> greg
>
>