> > ... should we revel
> > in configurability, and allow CREATE TABLE/ALTER TABLE behavior to
> > vary depending on the current threshold setting? We'd have to fix
> > toaster routines to not try to push stuff out-of-line when there is
> > out-of-line to push to ... but I think we probably had better do
> > anyway for robustness, if we're allowing any variability at all in
> > these numbers.
> Actually, upon looking closely at the toast code, it already
> does the right thing when there's no toast table. Good on
> someone for getting that right. But we still need to think
> about whether it's sane for CREATE/ALTER TABLE to condition
> the create-a-toast-table decision on a parameter that may now
> be changeable.
I think it is ok to decide during creation with current settings.
When a user wants a toast table that has not been created we can direct
them to use some dummy "alter table ... set storage ..." and create a
table if it does not exist (and the new settings opt for one).
And a new threshold has immediate consequences for inline compression,
so a change is not ignored.
In response to
pgsql-hackers by date
|Next:||From: Heikki Linnakangas||Date: 2007-04-03 08:56:00|
|Subject: Re: CheckpointStartLock starvation|
|Previous:||From: Simon Riggs||Date: 2007-04-03 07:55:56|
|Subject: Re: Interaction of PITR backups andBulkoperationsavoiding WAL|